Census

Summary of Important Changes

1. No substantive change to the Master Subscription Agreement

The contractual text of the Master Subscription Agreement appears unchanged. The diff primarily reflects:

• Website navigation and footer restructuring.
• Rebranding or migration from a Census-branded legal page to a Fivetran legal-site structure.
• Addition of links to Fivetran-specific materials, including:
• “Requirements for Fivetran AI”
• “Requirements for Free Plan”
• “Acceptable Use Policy”
• “Privacy Notice”
• “Website Terms of Use”
• “Privacy Program” and related compliance materials.
• Removal or alteration of page-level items such as “Trust Center,” title/date metadata, and “Need help?” formatting.

These changes do not themselves amend the operative agreement unless the linked materials are incorporated into an Order Form, SOW, or otherwise made contractually binding.

2. AI training and use of Customer Data

No express new AI-training right identified

The diff does not add language expressly allowing Census/Fivetran to use Customer Data to:

• Train, fine-tune, or improve artificial intelligence or machine-learning models;
• Develop generalized models;
• Create derivative datasets or commercial products from Customer Data; or
• Retain Customer Data for model-training purposes after termination.

The existing Section 4.2 still grants only a limited, non-exclusive right to use Customer Data “as reasonably necessary” to provide the Services. Section 7.2 also limits use of Confidential Information to purposes within the scope of the Agreement.

AI-related risk requiring further review

The navigation now links to “Requirements for Fivetran AI.” That document is not included in the diff. It could contain separate terms governing AI features, including data retention, human review, model improvement, or training. It should be reviewed to determine whether it is incorporated by reference or applies when AI features are used.

The agreement’s document-priority clause places the Order Form and SOW above the Agreement, and the Documentation below it. Any AI-specific terms in an Order Form or SOW could therefore alter the analysis.

3. Existing data-related risks remain

• Customer authorizes processing and transfers of Customer Data through connected Third-Party Services.
• Census/Fivetran is not responsible for a Third-Party Service’s access to, use, modification, or deletion of Customer Data.
• Customer indemnifies Census/Fivetran for claims concerning Customer Data when used as permitted.
• Customer remains responsible for obtaining required consents and complying with applicable law.
• The agreement does not provide detailed deletion, return, or post-termination data-handling procedures in the text shown.

Bottom line

No contractual expansion of AI-training rights is visible in this diff. However, the newly linked Fivetran AI terms should be reviewed before using AI features, because they may contain additional data-use permissions not present in the Master Subscription Agreement.

Summary

Key Finding

The diff only states that approximately 3,985 words were removed. It does not identify:

• Which clauses were deleted;
• Whether any provisions were added or replaced;
• The subject matter of the removed language; or
• Whether the deletions affect customer data, artificial intelligence, confidentiality, security, ownership, or liability.

Accordingly, the legal impact cannot be reliably determined from the information provided.

Potential Risks

A deletion of this size could materially change the agreement. Important provisions may have been removed concerning:

• AI model training: restrictions or permissions to use customer data, prompts, outputs, or usage data to train, fine-tune, evaluate, or improve AI models;
• Data ownership and licensing: the customer’s ownership of data and the provider’s rights to access, copy, analyze, or create derivative works;
• Confidentiality: protections applicable to customer data and exceptions allowing broader use;
• Privacy and security: data-processing obligations, security standards, breach notification, deletion, or data-retention requirements;
• Subprocessors and third parties: permission to disclose or transfer data to affiliates, vendors, or model providers;
• Liability and indemnification: limitations or exclusions relating to data misuse, confidentiality breaches, or AI-generated content;
• Termination and deletion: obligations to return or delete customer data and whether data may remain in model weights, logs, backups, or evaluation datasets; and
• Audit and compliance rights: customer rights to verify the provider’s use and protection of data.

AI-Training Assessment

No specific conclusion can be reached regarding AI-model training because the diff contains no operative language. In particular, it is not possible to determine whether the deletion:

1. Removes a prohibition on using customer data for AI training;

2. Removes the customer’s consent to such use;

3. Removes an opt-out or exclusion for customer data;

4. Removes restrictions on using prompts, outputs, or telemetry; or

5. Removes obligations to delete training data or prevent customer data from being retained in model parameters.

Information Needed

To complete the review, provide the actual redline or the deleted and revised clauses. At minimum, the review should include sections addressing:

• Customer data and permitted use;
• AI or machine-learning functionality;
• Confidentiality and privacy;
• Data retention and deletion;
• Subprocessors and disclosures;
• Intellectual property; and
• Liability and indemnification.

A deletion-only summary is insufficient to identify the specific legal changes or associated risks.

Structured Summary of Important Changes

1. Document and Website Navigation Changes

The diff primarily changes surrounding website navigation and branding:

• Adds or restores links for:
• Fivetran HVR Requirements
• Hybrid Deployment Requirements
• Free Plan Requirements
• Fivetran AI Requirements
• dbt Labs Requirements
• Acceptable Use Policy
• Other privacy, ethics, compliance, and website-term pages.
• Changes website labels and navigation, including:
• “Trust Center” formatting and surrounding page controls.
• “Website terms of use” and “Privacy policy” links.
• Copyright notices and footer content.
• The navigation appears to shift between Fivetran-branded legal pages and a Census Master Subscription Agreement.

These changes may affect where customers locate applicable product-specific terms, but they do not themselves modify the contractual rights or obligations in the Census agreement.

2. Substantive Agreement Terms

No substantive changes to the operative Census Master Subscription Agreement are apparent in the provided diff. The agreement continues to provide that:

• Customer owns its Customer Data.
• Census may use Customer Data only as “reasonably necessary” to provide the Services.
• Customer authorizes Census to transfer, store, process, and export Customer Data through connected third-party services.
• Customer Data is treated as Customer Confidential Information.
• Census may use customer names, logos, and marks for marketing.
• Customer indemnifies Census for claims relating to Customer Data when used as permitted by the agreement.
• Liability is generally capped at fees paid or payable during the preceding 12 months, subject to stated exclusions.

3. AI Model Training

No Express Training Authorization Identified

The diff does not add language expressly authorizing Census or Fivetran to:

• Train artificial intelligence or machine-learning models using Customer Data;
• Use Customer Data to improve, develop, or fine-tune models;
• Retain Customer Data for model-training purposes;
• Create or commercialize models derived from Customer Data; or
• Use Customer Data for product development beyond providing the Services.

The phrase allowing use of Customer Data “as reasonably necessary for Census to provide the Services” remains unchanged. Depending on the nature of an AI-enabled service, that wording could potentially be argued to cover limited processing needed to operate the service, but it does not clearly authorize general-purpose model training.

4. Risks and Points for Confirmation

• The navigation now references separate “Requirements for Fivetran AI” terms. Those terms are not included in the operative agreement diff and may contain additional AI-data-use provisions.
• Customers using any AI-related product should review the separate Fivetran AI terms and the Data Processing Addendum.
• The privacy policy is incorporated by reference for data handling, so changes to that policy could be relevant even though the agreement itself contains no express AI-training provision.
• The agreement does not expressly prohibit use of Customer Data for AI training, creating some residual ambiguity if AI functionality is added later.

Summary of Changes

Key Limitation

The provided diff states only that approximately 3,985 words were removed. It does not identify which provisions were deleted or provide replacement language. As a result, the specific legal and commercial effects cannot be determined reliably.

Potentially Important Risks

Deleting a substantial portion of a contract may remove provisions relating to:

• Customer data ownership and permitted uses
• Confidentiality and security obligations
• Data retention, deletion, and return requirements
• Subprocessors and third-party service providers
• Intellectual property rights and license grants
• Warranties, disclaimers, indemnities, and liability limits
• Audit, compliance, and regulatory obligations
• Termination rights and post-termination data handling
• Service levels, support, and remedies
• Governing law and dispute resolution

The deletion could therefore materially change the parties’ rights and obligations, even if no new language was added.

AI Model Training and Data Use

The supplied diff does not show whether any provisions concerning AI training were added, deleted, or modified. It is therefore impossible to determine whether:

• Customer data may be used to train, fine-tune, or improve AI models
• Customer prompts, inputs, outputs, or usage data may be retained for model development
• Such use is limited to aggregated, anonymized, or de-identified data
• The customer must opt in or may opt out
• Human review or third-party model-provider access is permitted
• Customer data is excluded from training by default
• Any trained models, derived data, or improvements belong to the provider or customer
• Data is deleted from training systems after termination

Recommended Review

Obtain the full before-and-after text or a marked copy identifying the deleted provisions. Particular attention should be given to any deleted language containing terms such as “train,” “improve,” “develop,” “machine learning,” “artificial intelligence,” “models,” “inputs,” “outputs,” “usage data,” “de-identified,” “aggregated,” “retain,” or “service improvement.”

Until the deleted text is reviewed, the safest conclusion is that there may be unknown reductions in customer protections, including potentially significant changes to data-use and AI-training rights.

Summary of Important Changes

1. Document and branding changes

• The document appears to have been migrated from a Census/Sutro Labs agreement into the Fivetran legal-site framework.
• Navigation and footer content now reference Fivetran, including:
• Fivetran HVR and Hybrid Deployment requirements
• Free Plan, Fivetran AI, and dbt Labs requirements
• Fivetran Acceptable Use Policy and other legal materials
• Fivetran contact information and copyright notice
• However, the operative agreement still identifies the provider as “Sutro Labs Inc. d/b/a Census” and continues to use “Census” throughout. This creates potential uncertainty about:
• Which legal entity is the contracting party
• Whether Fivetran has assumed or assigned the agreement
• Which linked policies and product-specific terms apply

2. Customer data use

• No substantive change is shown to the core Customer Data provisions:
• Customer retains ownership of Customer Data.
• Customer grants Census a limited, non-exclusive right to use Customer Data “as reasonably necessary” to provide the Services.
• Census may transfer, store, process, and dispose of Customer Data under its privacy policy.
• Customer authorizes data transfers to and from connected Third-Party Services.
• The agreement continues to classify Customer Data as the customer’s Confidential Information and restricts use of confidential information outside the agreement’s scope.
• No express authorization to use Customer Data to train AI models is added. There is also no express prohibition, detailed definition of “training,” restriction on model improvement, or commitment that Customer Data will be excluded from generalized or foundation-model training.
• The new navigation reference to “Requirements for Fivetran AI” may signal separate AI terms, but those terms are not included in the operative text provided. Customers using AI features should review those terms separately.

3. New or continuing risks

• The “reasonably necessary” services-use license may be interpreted broadly unless the applicable Data Processing Addendum or AI-specific terms impose tighter limits.
• The privacy policy is incorporated by reference but is not included; it may contain additional data-use, retention, international transfer, or AI-related permissions.
• Customer indemnifies Census for claims relating to Customer Data when used as permitted, potentially shifting risk to the customer for data supplied to AI or other features.
• Census may use Suggestions for any lawful purpose, including incorporating or sublicensing them. Feedback could therefore be used for product or model development, although Suggestions are not expressly defined as Customer Data.
• The provider may change Services and terminate for suspension-related reasons, with limited contractual protection if functionality or AI behavior changes.

Recommended action

Confirm the contracting entity and obtain the current Fivetran AI terms, privacy policy, and DPA. Seek an express statement that Customer Data, prompts, outputs, and personal information will not be used to train generalized AI models without the customer’s prior written consent.

Summary of Important Changes

1. Major branding and entity changes

• The document changes from a Fivetran-branded legal page to a Census-branded agreement.
• References to Fivetran Inc. and Fivetran legal/contact information are replaced with:
• Sutro Labs Inc. d/b/a Census
• Census website and email addresses, including legal@getcensus.com
• The navigation and related legal-document links are also changed from Fivetran materials to Census materials.
• The footer date and copyright references change, including references to 2025/2026.

Risk: Confirm that the contracting entity is correctly identified and has authority to provide the services. The change from Fivetran to Census could reflect a migration, rebranding, or use of a different legal entity. This may affect ownership, support obligations, data-processing responsibility, and enforcement rights.

2. Customer Data rights and permitted use

The agreement states that:

• Customer retains ownership of Customer Data.
• Customer grants Census a limited, non-exclusive right to use Customer Data as reasonably necessary to provide the Services.
• Customer Data includes information submitted by Customer or End Users, including data imported from authorized third-party services.
• Census may transfer, store, process, and dispose of Customer Data in accordance with its privacy policy.
• Customer authorizes Census to export Customer Data to connected third-party services.

Risk: The agreement incorporates the privacy policy by reference, but does not provide detailed data-use restrictions, retention periods, deletion commitments, or specific limits on secondary use. The privacy policy may therefore be important to determining how data is handled.

3. AI model training

• No express provision authorizes Census to use Customer Data to train, fine-tune, evaluate, or improve AI models.
• There is also no express prohibition on such use.
• The license is limited to use “as reasonably necessary” to provide the Services. Whether model training is necessary to provide the Services could be disputed, particularly for AI-enabled features.
• The agreement’s broad treatment of “Suggestions” permits Census to use feedback for any lawful purpose, but Suggestions are distinct from Customer Data and this clause does not expressly authorize training on Customer Data.

Risk: Customers seeking a clear “no training” commitment should request language expressly stating that Customer Data, prompts, outputs, and confidential information will not be used to train or improve generalized AI models, except with written consent. Any Fivetran AI/Census AI-specific terms should also be reviewed because they may contain additional permissions or restrictions.

4. Other material risk allocation

• Census may modify the Services and may suspend access for security, liability, or breach concerns.
• Services, outputs, and beta services are largely provided “as is.”
• Liability is generally capped at fees paid or payable during the preceding 12 months, with broad exclusions for indirect damages and data loss.
• Customer indemnifies Census for claims involving Customer Data when used as permitted.
• Automatic renewal applies unless notice is given at least 30 days before expiration.
• Disputes generally require binding arbitration in San Francisco, California.

Summary of Important Changes

1. Substantive Agreement Terms

No substantive changes to the operative Master Subscription Agreement were identified. The provisions governing services, Customer Data, confidentiality, intellectual property, liability, indemnification, termination, and disputes appear unchanged in the supplied diff.

2. Customer Data and AI Model Training

• No new authorization to use Customer Data for AI training was added.
• The existing data-use language remains limited to:
• Section 2.2: transferring, storing, processing, and disposing of Customer Data under the privacy policy.
• Section 3.6: transferring Customer Data to and from authorized Third-Party Services.
• Section 4.2: using Customer Data as “reasonably necessary” to provide the Services.
• The Agreement does not expressly state whether Customer Data may or may not be used to:
• train, fine-tune, or improve artificial-intelligence or machine-learning models;
• create embeddings, derived datasets, or aggregated training data;
• permit third-party AI providers to use Customer Data for model development.
• Accordingly, the diff does not create a new AI-training right, but the existing broad “reasonably necessary” service-use license and privacy-policy incorporation may leave ambiguity if AI functionality is introduced later. Customers requiring a prohibition on model training should seek express language stating that Customer Data, prompts, outputs, and usage data will not be used to train or improve models except with specific written consent.

3. Website Navigation and Branding Changes

The main changes concern the website’s legal-navigation content:

• The navigation now uses Fivetran-branded categories, including:
• “Requirements for HVR”
• “Requirements for Hybrid Deployment”
• “Requirements for Free Plan”
• “Requirements for Fivetran AI”
• “Requirements for dbt labs”
• “Acceptable Use Policy”
• Privacy-related links were added or reorganized, including “Privacy Notice,” “Global Applicant Privacy Notice,” “Cookie List,” and “Privacy Program.”
• Footer and page-navigation elements were updated from a 2025 Trust Center presentation to a 2026 Fivetran presentation, including “Go back to fivetran.com.”

These changes appear editorial or structural and do not themselves amend the Agreement unless incorporated by reference or otherwise made contractually binding.

4. Potential Documentation and Contracting Risks

• The document continues to identify the contracting provider as “Sutro Labs Inc. d/b/a Census,” while the surrounding website materials identify Fivetran Inc. This branding/entity mismatch could create uncertainty about the correct contracting party, applicable policies, and enforcement rights.
• The navigation references additional AI and service-specific requirements, but those documents are not substantively included in the diff. Customers should confirm whether an applicable “Fivetran AI” policy or other linked document contains separate data-use or AI-training permissions.
• Formatting artifacts and duplicated page content should be corrected to avoid uncertainty about which website terms or policies are intended to apply.

Executive Summary

The diff appears to be primarily a website migration, rebranding, and navigation update rather than a substantive amendment to the Master Subscription Agreement. The agreement continues to identify Sutro Labs Inc. d/b/a Census as the provider, although surrounding website content refers to Fivetran. This creates potential ambiguity about which entity, website terms, and policies apply.

Important Changes and Risks

1. Website and branding changes

• The legal navigation changes from Census-related content to a Fivetran-branded legal center.
• The page footer and links now refer to Fivetran Inc., Fivetran website terms, and Fivetran privacy policy.
• The agreement itself still names Sutro Labs Inc. d/b/a Census and continues to use Census-specific email addresses and URLs.

Risk: The inconsistent branding and entity references could create uncertainty regarding:

• The contracting party;
• The applicable privacy policy and other incorporated documents;
• Where legal notices must be sent; and
• Whether Fivetran has assumed or acquired Census’s contractual rights and obligations.

The assignment clause permits assignment to an affiliate or in connection with a merger or sale of substantially all assets, but the diff does not expressly document such a transfer.

2. Customer Data use

No substantive changes were identified to the Customer Data provisions:

• Customer retains ownership of Customer Data.
• Census receives a limited, non-exclusive license to use Customer Data “as reasonably necessary” to provide the Services.
• Census may transfer, store, and process Customer Data for data exchange with authorized third-party services.
• Customer remains responsible for obtaining necessary consents and complying with applicable law.

3. AI model training

The diff does not add an express provision authorizing or prohibiting the use of Customer Data to train, fine-tune, evaluate, or improve AI models.

Accordingly:

• There is no clear contractual commitment that Customer Data will not be used for AI training.
• The existing “reasonably necessary to provide the Services” license arguably does not clearly authorize generalized model training, but it could leave room for dispute depending on how AI functionality is characterized.
• The definition of Customer Data is broad and includes data obtained from third-party services, increasing the potential scope of any ambiguity.
• The agreement incorporates the provider’s privacy policy by URL. If that policy changes or is replaced by a Fivetran policy containing AI-related permissions, the practical data-use position could change without an amendment to the MSA.

Recommended action: Obtain written confirmation that Customer Data, prompts, outputs, and connected third-party data will not be used to train or improve generalized AI models, unless expressly authorized by Customer in writing. The confirmation should also address de-identified, aggregated, telemetry, and usage data.

4. Other substantive terms

No material changes were apparent to:

• Fees, renewals, termination, confidentiality, indemnities, liability caps, arbitration, or governing law;
• Customer ownership of data;
• Security commitments, including SOC 2 Type II; or
• Restrictions on use of the Services.

Summary of Important Changes

1. No substantive changes to the Master Subscription Agreement

The operative contract text appears unchanged. The diff primarily reflects:

• Website navigation and page-template changes.
• Reordering and renaming of links in the legal-home menu.
• Updated footer and branding references, including:
• “Fivetran” navigation and footer content.
• Updated copyright year from 2025 to 2026.
• Changes to “Trust Center” page formatting.
• Addition or reorganization of links such as “Fivetran HVR,” “Hybrid Deployment,” “Free Plan,” “Fivetran AI,” and “dbt labs.”
• Minor apparent typographical or formatting artifacts, such as “RequirementsRequirements,” “ee.g.,” and “Confidential Information t.”

These changes do not appear to amend the contractual rights or obligations in the Master Subscription Agreement.

2. Customer data use and AI training

No express AI-training authorization added

The diff does not add language authorizing Census/Fivetran to:

• Train artificial intelligence or machine-learning models using Customer Data.
• Improve generalized or shared AI models using Customer Data.
• Retain Customer Data for model training after termination.
• Use Customer Data to create derived datasets, embeddings, prompts, outputs, or other AI-related products.

The existing data-use provision remains:

> Customer grants Census a limited, non-exclusive right to use Customer Data as reasonably necessary to provide the Services.

This language is limited on its face to service provision and does not expressly permit model training. However, it does not define whether AI-related processing performed as part of a Service constitutes “providing the Services.”

AI-related navigation is not an operative amendment

The addition of a navigation link for “Requirements for Fivetran AI” may indicate that a separate AI-specific agreement or product terms exist. The navigation change alone does not incorporate those terms into this Agreement. Customers should nevertheless review the linked AI requirements, Order Forms, Documentation, and Data Processing Addendum to determine whether they contain separate AI-training permissions or restrictions.

3. Continuing data-related risks

The unchanged provisions continue to create these risks:

• Census may process Customer Data through Third-Party Services as authorized by the customer.
• Census disclaims responsibility for Third-Party Services’ access to, use, modification, or deletion of Customer Data.
• Customer remains responsible for obtaining required consents and complying with applicable law.
• Census’ privacy policy is incorporated by reference for the transfer, storage, processing, and disposal of Customer Data and may contain additional data-use rights.
• Customer indemnifies Census for claims relating to Customer Data when used as permitted under the Agreement.

Overall assessment

No material contractual change or new express AI-training right is visible in this diff. The principal action item is to review the separately linked AI-specific terms and privacy policy, because those documents may address AI data use outside the unchanged Master Subscription Agreement.

Summary of Material Changes

Scope of the Diff

The diff states that approximately 3,985 words were removed, but it does not identify which clauses, sections, or language were deleted. Because the deleted text is not provided, a reliable clause-by-clause legal analysis is not possible.

AI Training and Data Use

• The diff does not expressly show whether any provisions concerning the customer’s data being used to train AI models were added, deleted, or modified.
• However, the removal of approximately 3,985 words creates a significant risk that important data-use terms may have been eliminated, including:
• Restrictions or permissions to use customer data for AI model training, testing, or improvement;
• Definitions of “Customer Data,” “Usage Data,” “Content,” or de-identified data;
• Requirements to obtain customer consent;
• Opt-out rights or limitations on training use;
• Commitments not to use customer data to train generalized or third-party models;
• Data-retention, deletion, or anonymization requirements;
• Confidentiality and security protections;
• Ownership and intellectual-property provisions;
• Disclosure of subprocessors or third-party AI providers.

Key Legal Risks

1. Loss of express restrictions: If limitations on AI training were deleted, the provider may have broader discretion to use customer data than before.

2. Loss of consent or opt-out mechanisms: Deleted language may have removed customer approval requirements or the ability to opt out of model training.

3. Ambiguity: If related definitions or exceptions were removed, it may be unclear whether customer prompts, outputs, uploaded materials, metadata, or usage analytics can be used for training.

4. Reduced remedies and protections: Deleted confidentiality, security, indemnity, audit, or deletion provisions could increase the customer’s exposure if data is misused or disclosed.

5. Interpretation risk: The remaining agreement may contain gaps or conflicting language, potentially allowing the provider to argue for broader data-use rights.

Recommended Review

The full pre-change and post-change versions should be compared to identify the deleted sections. Particular attention should be given to provisions titled Data Use, Customer Data, Artificial Intelligence, Machine Learning, Confidentiality, Privacy, Security, Intellectual Property, Subprocessors, Retention, and Deletion. Until that comparison is available, the effect of the deletions—especially on AI training rights—should be treated as unresolved and potentially material.

Diff Analysis

Key Limitation

The supplied diff contains only the statement:

> “Added approximately 3985 words to the document”

It does not include the actual added, deleted, or replaced legal language. Accordingly, it is not possible to determine:

• What contractual provisions changed;
• Whether the changes expand or restrict either party’s rights;
• Whether liability, indemnification, confidentiality, termination, fees, or governing law changed; or
• Whether customer data may be used to train, fine-tune, evaluate, or improve AI models.

AI Training and Data-Use Risks

No conclusion can be reached regarding AI-model training because the relevant text is missing. The added language should be reviewed specifically for provisions addressing:

• Use of customer content, prompts, inputs, outputs, or account data to train or improve models;
• Whether use is automatic, opt-in, or opt-out;
• Use of data for “service improvement,” “product development,” analytics, research, or benchmarking;
• Whether data is aggregated, anonymized, de-identified, or retained in identifiable form;
• Human review or access to customer data;
• Sharing data with affiliates, vendors, subprocessors, or model providers;
• Retention and deletion periods;
• Customer ownership and provider license rights;
• Restrictions on using confidential, personal, regulated, or sensitive information;
• Commitments regarding model memorization, data leakage, or re-identification; and
• Whether enterprise, API, or paid customers receive different protections.

Information Needed

Please provide the actual redlined language, including the text inside:

• {additions};
• [deletions]; and
• []{replacements}.

If the document is too long, provide at least the newly added sections and any provisions concerning data, privacy, confidentiality, artificial intelligence, machine learning, model training, service improvement, content, inputs, outputs, or retention.

Summary of Important Changes

1. Document and Website Rebranding

• The document has been substantially replaced with a Census Master Subscription Agreement, identifying the provider as Sutro Labs Inc. d/b/a Census.
• The prior navigation and website content referred to Fivetran, including Fivetran-specific requirements, privacy materials, and legal pages.
• The revised website navigation now includes Census content but still contains extensive Fivetran branding, links, copyright notices, and legal-home elements.

Risk

This creates uncertainty about:

• Which entity is actually contracting with the customer;
• Which privacy policy, data-processing addendum, and product-specific terms apply;
• Whether the document was published or assembled correctly; and
• Whether Fivetran or Census may exercise contractual rights.

The customer should confirm the contracting entity and obtain a clean, internally consistent version of the agreement.

2. Customer Data Rights and Use

The revised agreement defines Customer Data broadly to include data submitted by the customer or its End Users, including data imported from authorized third-party services.

The customer retains ownership of Customer Data under Section 4.1. However, Section 4.2 grants Census a license to use Customer Data:

> “as reasonably necessary for Census to provide the Services.”

The agreement also authorizes Census to transfer, store, process, and export Customer Data through connected third-party services.

AI Training Analysis

There is no express new provision authorizing Census to use Customer Data to train, fine-tune, improve, or develop AI models. There is also no express prohibition on such use.

However, the following provisions create residual risk:

• The “provide the Services” license is not expressly limited to processing strictly necessary for a particular customer transaction.
• “Services” is defined broadly and may include future products or features.
• Census may process Customer Data under its privacy policy, which is incorporated by reference but not included in the diff.
• Beta Services may have different security standards and could potentially include AI functionality.
• Suggestions may be used by Census for any lawful purpose, although Suggestions are distinct from Customer Data.

Customers handling confidential, personal, regulated, or commercially sensitive information should seek an express contractual statement that Customer Data will not be used to train or improve generalized AI models, except with prior written consent, and that any AI processing will be disclosed and governed by the DPA.

3. Privacy and Security Changes

Census promises an industry-standard security program meeting SOC 2 Type II requirements. Data processing is also made subject to Census’s privacy policy.

Risk

The agreement does not itself specify:

• Data retention or deletion periods;
• Subprocessor controls;
• Data-location or cross-border-transfer restrictions;
• Customer audit rights; or
• Specific AI governance or model-security obligations.

These matters may be addressed in the separate Data Processing Addendum, which should be reviewed together with this agreement.

4. Liability and Indemnity

The customer indemnifies Census for claims relating to Customer Data when used as permitted under the agreement. Census’s liability is generally capped at fees paid or payable during the preceding 12 months, while indirect damages and data-loss damages are broadly excluded.

Risk

There is no specific uncapped liability protection for unauthorized AI training, misuse of Customer Data, or confidentiality breaches. The customer may therefore have limited recovery if data is used beyond permitted purposes.

5. Other Material Customer Risks

• Automatic subscription renewal unless notice is given at least 30 days before expiration.
• Fees are generally non-cancelable and non-refundable.
• Census may suspend or terminate service for broad security, liability, breach, or legal-compliance reasons.
• Binding arbitration in San Francisco, California.
• Census may use the customer’s name and logo in marketing materials.
• Beta Services may have reduced functionality, reliability, and security protections.

Summary

The provided diff does not include the actual contractual language. It only states:

> “Added approximately 3985 words to the document”

Accordingly, it is not possible to identify:

• Which provisions were added, deleted, or replaced;
• Changes to customer rights or obligations;
• New liability, indemnity, confidentiality, security, or compliance risks;
• Any changes to data ownership or permitted data use; or
• Whether customer data may be used to train, fine-tune, evaluate, or improve AI models.

AI Training and Data Use

No substantive AI- or data-use language is included in the diff. Therefore, the available information does not establish whether the contract:

• Permits using customer content, prompts, inputs, outputs, or personal data to train AI models;
• Allows use for model improvement, analytics, benchmarking, or product development;
• Applies such use by default or only with customer consent;
• Provides an opt-out or requires an opt-in;
• Uses data in de-identified, aggregated, or identifiable form;
• Allows disclosure to model providers, affiliates, subprocessors, or other third parties;
• Requires deletion or retention limits; or
• Grants the customer ownership or control over models trained using customer data.

Risk Assessment

The only identifiable fact is that approximately 3,985 words were added. The size of the addition suggests potentially material changes, but its legal effect cannot be assessed without the added and surrounding text.

Information Needed

Please provide the full redlined language, including:

1. All additions, deletions, and replacements;

2. The surrounding provisions needed to understand each change; and

3. Any definitions referenced by the new language.

Once provided, the changes can be analyzed for specific risks, including AI training permissions and customer-data protections.

Structured Summary of Important Changes

1. Major Document Replacement and Party Change

• The prior Fivetran legal-site content appears to have been replaced with a Master Subscription Agreement for Sutro Labs Inc. d/b/a Census.
• References to Fivetran, including Fivetran’s legal pages, privacy policy, and contact information, are replaced with Census branding and contact details.
• The footer and copyright references nevertheless retain some Fivetran references, creating potential entity-identification and document-version ambiguity.
• Customers should confirm that the correct contracting entity, governing terms, and privacy documents are incorporated into their Order Form.

2. Customer Data Rights and Use

The new agreement defines “Customer Data” broadly to include:

• Data submitted by the customer or its End Users; and
• Data imported from authorized Third-Party Services.

The customer retains ownership of Customer Data, but grants Census a limited, non-exclusive license to use it “as reasonably necessary” to provide the Services.

Census may transfer, store, process, export, and dispose of Customer Data through the Services and connected Third-Party Services. Census disclaims responsibility for Third-Party Services’ access to, use, modification, or deletion of Customer Data.

AI-model training

• No express authorization was identified allowing Census to use Customer Data to train, fine-tune, or improve AI models.
• The license is expressly tied to providing the Services, which weighs against a broad independent right to use Customer Data for model training.
• However, the agreement does not expressly state that Customer Data will not be used for AI training, nor does it address:
• De-identification or aggregation;
• Use of prompts, outputs, or telemetry;
• Whether subprocessors or AI providers may train on the data;
• Retention of data in model weights; or
• Customer opt-out or deletion rights.
• The broad definition of Customer Data and the incorporation of the external privacy policy create residual uncertainty. That policy should be reviewed for AI-related permissions.

3. Confidentiality and Privacy

• Customer Data is expressly classified as Customer Confidential Information.
• Census must use reasonable care and generally may not use confidential information outside the agreement’s scope.
• Census may disclose information to employees, contractors, affiliates, and Third-Party Services with a need to know.
• The agreement relies on Census’s online privacy policy for data handling rather than providing detailed data-processing terms in the agreement itself.
• There is no detailed deletion, return, breach-notification, data-location, subprocessors, or data-subject-rights framework in the text provided.

4. Customer and Commercial Risks

• Free accounts are limited to one per organization, and use on behalf of third parties generally requires an Embedded subscription.
• Subscriptions automatically renew unless notice is provided at least 30 days before expiration.
• Fees are generally non-cancelable and non-refundable; late fees may be 1.5% per month.
• Census may suspend service for security concerns, potential liability, or breach, and may terminate where it has a suspension right.
• Services and especially Beta Services are provided largely “as is,” with limited warranties and no warranty for Third-Party Services.
• Liability is generally capped at fees paid or payable in the preceding 12 months, while the customer’s indemnity obligations and breach of restrictions are excluded from the cap.
• Disputes are subject to binding AAA arbitration in San Francisco, California, with California law governing.

Summary

The provided diff only states: “Added approximately 3985 words to the document.” It does not include the actual added, deleted, or replaced legal language.

AI Training and Customer Data

• It is not possible to determine whether the changes:
• Permit the provider to use customer data to train, fine-tune, or improve AI models;
• Restrict training to de-identified, aggregated, or anonymized data;
• Allow use of prompts, inputs, outputs, uploaded files, metadata, or usage data;
• Apply different rules to consumer, business, or enterprise accounts;
• Permit human review or disclosure of customer data for model development;
• Allow retention of customer data after termination; or
• Give the customer an opt-out or require affirmative consent.

Other Legal Risks

The actual additions are also needed to assess potential changes concerning:

• Ownership and licensing of customer data and generated outputs;
• Confidentiality and data-security obligations;
• Privacy-law compliance and international data transfers;
• Service-provider or subcontractor access;
• Warranties, indemnities, and liability caps;
• Data retention, deletion, and termination rights;
• Intellectual-property rights in AI-generated material; and
• Changes to governing law, dispute resolution, or audit rights.

Required Information

Please provide the full diff, including the text inside {}, [], and []{}. The word-count statement alone does not identify any contractual changes or permit a reliable risk analysis.

Summary of Changes

Scope of the Diff

The diff states only that approximately 3,985 words were removed from the document. It does not identify:

• Which clauses were deleted;
• Whether any provisions were replaced or moved;
• The legal effect of the deletions; or
• Whether the deleted language concerned data use, confidentiality, security, intellectual property, or AI training.

AI Training and Customer Data

No specific change concerning the use of customer data to train AI models can be confirmed from the information provided.

However, deleting approximately 3,985 words could materially affect AI-related rights if the removed text included provisions addressing:

• Whether customer data may be used to train, fine-tune, test, or improve AI models;
• Whether prompts, inputs, outputs, or usage data are retained or reused;
• Whether data is used in aggregated, de-identified, or identifiable form;
• Whether the provider may use customer data for product development or analytics;
• Opt-out or consent requirements for AI training;
• Ownership of customer data, outputs, or model improvements;
• Confidentiality restrictions applicable to AI systems; or
• Deletion, retention, or segregation of data used for model training.

If any such restrictions were removed without replacement, the customer may face increased uncertainty or exposure to broader provider rights to use its data.

Other Potential Legal Risks

The deletion may also have removed protections relating to:

• Confidentiality and permitted disclosures;
• Data security and breach notification;
• Data processing roles and privacy-law compliance;
• Subprocessors and international data transfers;
• Indemnities and liability limitations;
• Audit, deletion, and termination rights; or
• Intellectual-property ownership and license grants.

The removal of liability, security, or confidentiality language could substantially increase the customer’s risk even if the commercial service remains unchanged.

Conclusion

The provided diff is insufficient to determine the actual legal changes. A reliable analysis requires the text of the deleted provisions and any surrounding or replacement language. In particular, the customer should verify whether the deleted material contained an express prohibition, limitation, or opt-out relating to the use of customer data for AI model training.

Diff Analysis

Key Limitation

The supplied diff only states:

> “Added approximately 3985 words to the document”

It does not include the text of the additions, deletions, or replacements. As a result, it is not possible to identify the legal changes, new obligations, allocation of risk, or changes concerning use of customer data to train AI models.

AI Training and Customer Data

No substantive language is provided addressing:

• Whether customer data may be used to train, fine-tune, test, or improve AI models;
• Whether prompts, inputs, outputs, files, or usage metadata may be retained;
• Whether customer data may be shared with affiliates, vendors, or model providers;
• Whether data is anonymized, aggregated, or de-identified before use;
• Whether customers can opt out of AI training or request deletion;
• Ownership or licensing rights in customer data, outputs, or model improvements;
• Confidentiality, security, or breach responsibilities relating to AI systems; or
• Whether customer data may be used to train models serving other customers.

Risk Assessment

Because the underlying text is missing, no reliable assessment can be made of:

• Expanded data-use rights;
• New customer warranties or indemnities;
• Changes to confidentiality obligations;
• Broader license grants;
• Reduced provider liability;
• New limitations of liability;
• Changes to termination or deletion rights; or
• Governing law and dispute-resolution provisions.

Information Needed

Please provide the actual redlined text, including the approximately 3,985 added words and any marked deletions or replacements. The complete diff is necessary to perform a meaningful legal review, particularly of provisions concerning customer data, artificial intelligence, confidentiality, intellectual property, security, and liability.

Structured Summary of Important Changes

1. Website and Branding Changes

The most significant visible changes are to the website’s legal-navigation and footer content:

• Fivetran-branded legal navigation has replaced or supplemented the prior Census-oriented presentation.
• New or reorganized links include:
• Requirements for Fivetran HVR
• Hybrid Deployment requirements
• Free Plan requirements
• Fivetran AI requirements
• dbt Labs requirements
• Acceptable Use Policy
• Privacy Notice and related privacy materials
• Footer and page metadata have changed, including references to:
• Fivetran Inc.
• Updated copyright years
• legal@fivetran.com
• “Website terms of use” and “Privacy policy”
• New page title/date/table-of-contents formatting

Risk/Issue

The MSA body continues to identify the contracting provider as Sutro Labs Inc. d/b/a Census, and continues to use legal@getcensus.com, while the surrounding website content refers to Fivetran and Fivetran legal contacts. This creates potential uncertainty regarding:

• Which legal entity is the contracting party;
• Whether Fivetran is an assignee, affiliate, successor, or merely the website operator;
• Which privacy policy and supplemental terms apply; and
• Where legal notices should be sent.

The agreement’s assignment clause permits Census to assign the agreement to an affiliate or in connection with a merger or sale, but the diff does not expressly document such an assignment.

2. Customer Data and AI Training

No Express AI-Training Change Identified

The substantive MSA language concerning Customer Data appears unchanged. In particular:

• “Customer Data” still includes data submitted by the customer, End Users, and authorized Third-Party Services.
• Customer retains ownership of Customer Data.
• Customer grants Census only a limited, non-exclusive right to use Customer Data as reasonably necessary to provide the Services.
• Confidentiality provisions restrict use of confidential information to purposes within the scope of the agreement.
• The MSA does not expressly authorize use of Customer Data to:
• Train, fine-tune, or improve artificial-intelligence or machine-learning models;
• Create generalized or de-identified datasets for model training; or
• Develop products unrelated to providing the Services.

Residual AI Risk

The navigation now includes “Requirements for Fivetran AI,” but the diff does not include the text of those requirements. Customers should review that separate document because it could contain AI-specific data-use permissions, disclaimers, retention terms, or restrictions incorporated through the Documentation or applicable Order Form.

The MSA also incorporates the provider’s privacy policy by reference, and that policy may address AI-related processing outside the MSA’s text. No conclusion about AI training should be made without reviewing the current privacy policy and Fivetran AI requirements.

3. Overall Assessment

The diff appears primarily to reflect a rebranding, site migration, and reorganization of legal materials, rather than substantive changes to the MSA’s operational terms. The principal legal concern is the apparent mismatch between the Census contracting entity and Fivetran-branded legal materials.

Summary

The provided diff does not include the actual contractual language. It only states:

> “Added approximately 3985 words to the document”

Accordingly, it is not possible to identify:

• Which provisions were added, deleted, or replaced;
• Changes to customer rights or provider obligations;
• New liability, confidentiality, security, or indemnity risks;
• Changes to data ownership, licensing, retention, or disclosure; or
• Whether customer data may be used to train, fine-tune, evaluate, or improve AI models.

AI Training and Data Use

No substantive language addressing AI model training or the use of customer data is included in the supplied diff. The statement that approximately 3,985 words were added does not establish whether the new terms:

• Permit training on customer data by default or only with consent;
• Apply to prompts, inputs, outputs, usage data, or derived data;
• Allow human review or sharing with affiliates and subprocessors;
• Permit retention after contract termination;
• Grant the provider ownership or a broad license;
• Require de-identification or anonymization; or
• Provide opt-out, deletion, audit, or security protections.

Information Needed

A meaningful legal review requires the full redline or the text of the additions, deletions, and replacements. The added clauses should be provided using the stated notation, or as a comparison showing the prior and revised versions.

Summary

Scope of the Change

The diff only states that approximately 3,985 words were removed. It does not identify which provisions were deleted or provide replacement language.

AI Training and Customer Data

• No determination is possible regarding whether the agreement now permits, restricts, or prohibits using customer data to train AI models.
• The deletion may have removed provisions addressing:
• Use of customer content, personal data, or confidential information for AI training;
• Whether data is used to train general-purpose or provider-specific models;
• Customer consent or opt-out rights;
• De-identification, aggregation, or anonymization requirements;
• Human review or model-improvement activities;
• Restrictions on using prompts, outputs, files, or usage metadata;
• Data retention and deletion after termination; or
• Allocation of intellectual-property, confidentiality, and privacy responsibilities.
• Without the deleted text and any replacement provisions, it is not possible to determine whether the customer’s position has improved or worsened.

Potential Legal Risks from the Deletion

Removing nearly 4,000 words could materially affect:

• Data-use restrictions: Deleted limits may allow broader use of customer data.
• Confidentiality: Deleted confidentiality language could reduce protections for business information.
• Privacy compliance: Deleted provisions may have addressed applicable data-protection laws, security measures, or processing instructions.
• Intellectual property: Deleted ownership or license terms could create uncertainty about customer content, outputs, or model improvements.
• Security and liability: Deleted warranties, incident-notification duties, indemnities, or liability protections may shift risk to the customer.
• Termination and deletion: Deleted obligations may weaken the customer’s ability to require deletion or prevent continued use of data.
• Transparency and control: Deleted audit, notice, opt-out, or reporting rights could limit the customer’s oversight.

Required Information for a Reliable Analysis

Please provide either:

1. The full redlined text showing the deleted and added provisions; or

2. The specific deleted and replacement clauses, particularly those addressing customer data, artificial intelligence, machine learning, model training, confidentiality, privacy, security, intellectual property, and data retention.

On the current information, the only definite conclusion is that a substantial amount of contractual language was removed; the legal and AI-training effects cannot be reliably assessed.

Key Changes and Risks

1. Primarily website/navigation changes

The diff largely changes the surrounding website navigation and footer content rather than the operative Master Subscription Agreement. Notable changes include:

• Addition of links or references to:
• “Requirements for Fivetran AI”
• “Requirements for dbt labs”
• “Free Plan”
• “Acceptable Use Policy”
• “Privacy Notice,” “Cookie List,” and other privacy resources
• Changes to navigation labels, including:
• “Requirements for Fivetran HVR”
• “Requirements for Hybrid Deployment”
• “Website Terms of Use”
• “Privacy”
• Updated branding and footer information, including “Fivetran Inc.” and a 2026 copyright notice.
• Removal or relocation of certain navigation text, such as references to “Fivetran Inc.”, “Website terms of use,” and “Privacy policy.”

These changes may affect which supplemental terms are accessible through the website, but they do not, by themselves, amend the substantive agreement unless those linked documents are incorporated or otherwise made applicable.

2. No apparent change to Customer Data rights

The operative data provisions appear unchanged:

• Customer Data definition: Still includes data submitted by Customer, End Users, and authorized Third-Party Services.
• Customer ownership: Customer and its licensors continue to own Customer Data.
• Census license: Census retains a limited, non-exclusive right to use Customer Data only “as reasonably necessary” to provide the Services.
• Confidentiality: Customer Data remains classified as Customer Confidential Information.
• Data exchange: Census remains authorized to transfer, store, process, and export Customer Data through connected Third-Party Services.
• Third-party responsibility: Census disclaims responsibility for Third-Party Services’ handling, modification, or deletion of Customer Data.

3. AI model training

No express provision was added or deleted that authorizes Census/Fivetran to:

• Train, fine-tune, or improve AI models using Customer Data;
• Use Customer Data to develop generalized models;
• Retain prompts, outputs, embeddings, or derivative data for AI training; or
• Prohibit AI training using Customer Data.

The new navigation reference to “Requirements for Fivetran AI” may indicate a separate AI-specific document. That document should be reviewed because it could contain additional data-use permissions or restrictions not present in this Agreement. The Agreement’s integration of external Documentation and linked policies also creates a risk that AI-related terms may apply outside the four corners of the displayed agreement.

4. Branding and entity-name risk

The agreement continues to identify the provider as “Sutro Labs Inc. d/b/a Census,” while the website navigation and footer refer to Fivetran and Fivetran Inc. This inconsistency may create uncertainty regarding:

• The contracting entity;
• Which privacy policy and supplemental requirements apply;
• Whether Fivetran has assumed or been assigned the agreement; and
• Which entity is responsible for data protection and contractual obligations.

This should be clarified in the Order Form or through an express amendment.

Executive Summary

The diff appears to replace a Fivetran-branded legal webpage with a Census-branded Master Subscription Agreement, while also adding links to several product-specific requirements, including “Requirements for Fivetran AI.” The body of the agreement is otherwise largely unchanged. The most important risks concern entity/brand inconsistency, incorporation of external policies, and the absence of clear AI-training restrictions.

Important Changes and Risks

1. Contracting-party and branding changes

• The agreement now identifies the provider as “Sutro Labs Inc. d/b/a Census” rather than Fivetran.
• Website navigation and footer references are inconsistent:
• New text refers to Census and getcensus.com.
• Other additions still refer to Fivetran, fivetran.com, legal@fivetran.com, and “Fivetran Inc.”
• The legal navigation includes both Census-related content and Fivetran-specific requirements.

Risk: It may be unclear which entity is the actual contracting party, which policies apply, and where notices must be sent. This should be corrected before execution or reliance.

2. New or surfaced product-specific requirements

The navigation adds or highlights requirements for:

• Fivetran HVR
• Hybrid Deployment
• Free Plan
• Fivetran AI
• dbt Labs
• Acceptable Use

Risk: These linked documents may impose additional usage restrictions, data rights, security terms, or AI-related conditions. Because the agreement incorporates Documentation and other attachments by reference, customers should obtain and review the applicable versions. The diff does not show their substantive text, so their legal effect cannot be determined from this comparison alone.

3. Customer Data use and AI model training

The agreement continues to define Customer Data broadly, including data submitted by End Users and connected third-party services.

The provider receives a license to use Customer Data:

> “as reasonably necessary for Census to provide the Services.”

The agreement does not expressly state that Customer Data may be used to:

• Train, fine-tune, or improve generative AI or machine-learning models;
• Create generalized models or datasets;
• Share data with AI vendors for model development; or
• Use Customer Data after termination for training.

Conversely, the confidentiality provisions restrict use of Confidential Information to purposes within the agreement’s scope. Customer Data is expressly classified as Customer Confidential Information.

Risk assessment: On the face of this agreement, there is no clear affirmative authorization for training AI models, and the limited service-provision license may not support such use. However, the newly linked Fivetran AI requirements, privacy policy, Documentation, or other incorporated materials could alter that conclusion. Customers should obtain a written statement that Customer Data, prompts, outputs, and derived data will not be used to train or improve models without express consent, and should address de-identification, retention, subprocessors, and deletion.

4. External privacy policy controls data handling

Section 2.2 requires processing in accordance with the provider’s privacy policy, available online.

Risk: The privacy policy can potentially change independently and may contain broader data-use permissions than the agreement. Customers should require contractual precedence for the agreement/DPA and review change-notice rights.

5. Other material commercial risks

• Automatic renewal unless notice is given at least 30 days before term end.
• Fees are generally non-cancelable and non-refundable.
• Broad suspension and termination rights for the provider.
• Liability cap generally limited to fees paid or payable in the prior 12 months.
• Data loss and service interruption damages are excluded.
• Customer indemnifies the provider for claims relating to Customer Data used as permitted.
• Mandatory arbitration in San Francisco, California.

Executive Summary

The diff appears primarily to update the website’s navigation, branding, and footer content from a Census-branded legal site to a Fivetran-branded legal site. The substantive Master Subscription Agreement text is otherwise unchanged.

Important Changes

1. Website navigation and branding

• “Software and Service Specific Requirements” is retained, while navigation items are reorganized.
• New or newly displayed links include:
• Requirements for Fivetran HVR
• Requirements for Hybrid Deployment
• Requirements for Free Plan
• Requirements for Fivetran AI
• Requirements for dbt labs
• Acceptable Use Policy
• Support Policy
• Service Level Agreement
• EU Data Act Addendum
• Additional privacy and compliance links are displayed, including:
• Privacy Notice
• Global Applicant Privacy Notice
• Cookie List
• Privacy Program
• Ethics and Compliance
• Code of Conduct
• Anti-Slavery and Human Trafficking Statement

Risk: The newly listed documents may contain terms that apply to the Services, particularly the Fivetran AI Requirements, Free Plan Requirements, Acceptable Use Policy, and EU Data Act Addendum. Their inclusion in navigation does not, by itself, establish whether they are incorporated into the Agreement or which document controls in the event of a conflict. The Agreement’s stated order of precedence remains: Order Form, SOW, Agreement, then Documentation.

2. Rebranding and entity inconsistency

• The page header, footer, and links are changed from Census/Fivetran legacy website references to updated Fivetran navigation.
• The Agreement itself continues to identify the provider as “Sutro Labs Inc. d/b/a Census” and repeatedly refers to “Census.”
• Footer details change to “© 2026 Fivetran Inc.”

Risk: The mixed Census/Fivetran branding could create uncertainty about the contracting entity, applicable policies, notice details, or whether the displayed terms are current. The agreement’s entity name and Order Form should be checked carefully.

AI Training and Customer Data

No express AI-training authorization appears in the diff

The diff does not add or remove language expressly authorizing, prohibiting, or regulating the use of Customer Data to train AI or machine-learning models.

The existing provisions state that:

• Customer retains ownership of Customer Data.
• Customer grants Census a limited, non-exclusive license to use Customer Data “as reasonably necessary” to provide the Services.
• Confidential Information, including Customer Data, may not be used outside the Agreement’s permitted scope.
• The customer must authorize Census to transfer, store, and process data through connected third-party services.

Risk: Although the current language does not clearly permit generalized model training, it also does not expressly prohibit training, model improvement, or use of de-identified or aggregated data. The newly linked Fivetran AI Requirements should be reviewed for any additional AI-data rights, especially if it is incorporated through Documentation or an Order Form.

Structured Summary of Important Changes

1. Document and Company Branding Changes

• The legal site navigation has been substantially changed from Fivetran materials to Census materials.
• References to Fivetran-specific documents and requirements were replaced with Census-related materials, including:
• “Requirements for Fivetran HVR”
• “Requirements for Hybrid Deployment”
• “Requirements for Free Plan”
• “Requirements for Fivetran AI”
• “Requirements for dbt labs”
• The agreement now identifies the provider as “Sutro Labs Inc. d/b/a Census” rather than Fivetran.
• Website, privacy-policy, legal-contact, and footer references were changed from Fivetran to Census, including:
• legal@fivetran.com → legal@getcensus.com
• Fivetran website/privacy links → Census website/privacy links
• Footer copyright references changed to Census.

Risk/impact: These changes may indicate a rebranding, acquisition, or document-template substitution. Customers should confirm that the contracting entity, privacy policy, DPA, security commitments, and applicable order forms are all consistent with the intended service provider. The navigation includes references to Fivetran documents that do not appear to form part of the agreement itself.

2. Customer Data Use

• No express provision was added authorizing Census to use Customer Data to train, fine-tune, evaluate, or improve AI models.
• The principal data-use grant remains limited to use of Customer Data “as reasonably necessary” to provide the Services under Section 4.2.
• Section 2.2 continues to state that Customer Data will be processed under Census’ privacy policy.
• Section 7 treats Customer Data as Customer Confidential Information and restricts use of confidential information to purposes within the agreement.
• Section 3.6 authorizes Census to transfer, store, process, and export Customer Data in connection with connected third-party services.

AI-specific risk assessment: The diff does not add a direct AI-training right. However, the “provide the Services” language could potentially encompass AI-enabled functionality if the Services include AI features. The agreement does not expressly address whether prompts, inputs, outputs, or Customer Data may be retained or used for model improvement, nor does it provide a clear opt-out, deletion commitment, model-training prohibition, or distinction between customer-specific and generalized data. Review the privacy policy, DPA, and any separate “Requirements for Fivetran AI” or Census AI terms.

3. Other Material Risk Points

• Customer indemnifies Census for claims concerning Customer Data when used as permitted by the agreement.
• Third-party services may access, modify, or delete Customer Data, with Census disclaiming responsibility for those actions.
• Liability exclusions cover loss of data and service interruption; the general liability cap is limited to fees paid or payable in the preceding 12 months, subject to specified exclusions.
• Automatic renewal, suspension, termination, arbitration in San Francisco, and California governing law remain unchanged in the apparent agreement text.

Bottom Line

The substantive diff appears primarily to replace Fivetran branding and references with Census branding. No new AI-training permission is visible, but the agreement lacks explicit AI-data safeguards, so the privacy policy, DPA, and any AI-specific terms should be reviewed before acceptance.

The publisher records this document as revised on this date (“Last updated: February 19, 2025”).

No

Summary of Changes

Overall Change

• The amendment removes approximately 3,985 words from the document.
• No replacement language or new terms is provided in the diff.
• Because the deleted text is not included, it is not possible to determine which obligations, rights, limitations, or protections were removed.

Customer Data and AI Training

• The diff does not identify whether any provisions concerning the use of customer data to train AI models were added, removed, or modified.
• However, the deletion of a substantial portion of the agreement creates a significant risk that provisions addressing AI may have been removed, including:
• Whether customer data may be used to train, fine-tune, or improve AI models;
• Whether customer prompts, inputs, outputs, or usage data may be retained for training;
• Whether data is anonymized, aggregated, or de-identified before such use;
• Whether the customer must opt in or may opt out of AI training;
• Whether the provider may use customer data to train models shared with other customers;
• Restrictions on using confidential, personal, or regulated data for training;
• Data deletion, retention, and audit obligations; and
• The provider’s responsibility for unauthorized disclosure or misuse of training data.

Key Legal Risks

• Loss of customer protections: Deleted confidentiality, data-security, privacy, or data-use restrictions could reduce the customer’s contractual protections.
• Expanded implied permissions: If a deleted restriction previously limited the provider’s use of data, its removal may leave broader rights under the remaining agreement or applicable law.
• Unclear data-use rights: The agreement may no longer clearly state whether customer data can be used for AI development or product improvement.
• Compliance exposure: Removing provisions concerning personal data, regulated information, or international transfers could create uncertainty regarding compliance responsibilities.
• Reduced remedies: Deleted liability, indemnity, audit, notice, or deletion provisions could make it more difficult for the customer to respond to misuse or a data breach.
• Interpretation risk: A large deletion without replacement language may create gaps, inconsistencies, or unintended reliance on general terms elsewhere in the agreement.

Recommended Review

The complete pre- and post-amendment versions should be compared, especially provisions concerning:

1. Customer data ownership and permitted uses;

2. AI training and model improvement;

3. Confidentiality and security;

4. Privacy and data-processing obligations;

5. Retention and deletion;

6. Subprocessors and data sharing;

7. Liability, indemnities, and remedies; and

8. Termination and post-termination data handling.

Based solely on the supplied diff, the principal conclusion is that a material amount of contractual language has been removed, but the specific legal and AI-related consequences cannot be determined without the deleted text or the full versions of the agreement.

Summary

The diff only states: “Added approximately 3985 words to the document.” It does not include the actual added contractual language.

Analysis Limitations

Because the substantive text is missing, it is not possible to determine:

• What contractual terms were added or changed.
• Whether liability, indemnity, confidentiality, security, privacy, intellectual-property, or termination provisions changed.
• Whether the customer’s data may be used to train, fine-tune, evaluate, or improve AI models.
• Whether data may be shared with affiliates, vendors, subprocessors, or model providers.
• Whether customer data is retained, anonymized, aggregated, or deleted.
• Whether the customer has opted out of AI training or has any objection, deletion, or audit rights.
• Whether the provider obtains rights to use customer inputs, outputs, prompts, files, or usage data.
• Whether the agreement distinguishes between customer content, personal data, metadata, and de-identified information.

Key Concern

The addition of nearly 4,000 words could materially expand the customer’s obligations or the provider’s rights, but the scope and legal effect cannot be assessed without the actual text. In particular, no conclusion can be reached regarding AI-model training or use of customer data.

Needed for Further Review

Please provide the full redlined language, including:

1. The deleted text in brackets.

2. The added text in braces.

3. Any replacement provisions shown using the specified notation.

4. Relevant surrounding sections if the additions modify existing definitions or cross-references.

Once provided, the changes can be analyzed for legal effect, commercial risk, and AI-data-use implications.