Vercel

Summary

The diff only states that approximately 76 words were removed from the document. The actual deleted language is not provided.

AI Training and Data Use

• It is not possible to determine whether the changes affect:
• The customer's permission for the provider to use customer data to train, fine-tune, or improve AI models;
• Whether customer data may be used for general-purpose or third-party models;
• Whether prompts, outputs, usage data, personal information, or confidential information are included;
• Any opt-out, consent, or limitation on AI training;
• Data retention, anonymization, aggregation, or deletion requirements; or
• Restrictions on using customer data to create derivative datasets or model improvements.

Potential Legal Significance

Removing language can either narrow or expand the customer's protections, depending on what was deleted. For example:

• Deleting a prohibition on model training could permit broader use of customer data.
• Deleting a consent or opt-out mechanism could reduce the customer's control.
• Deleting confidentiality, security, or deletion obligations could increase privacy and confidentiality risk.
• Deleting a provider authorization could restrict the provider's ability to use data for service improvement.

However, no specific legal conclusion can be reached without the actual 76 deleted words and their surrounding provisions.

Information Needed

To analyze the amendment reliably, provide:

1. The exact deleted text;

2. The provision or section where it appeared; and

3. Any remaining or replacement language.

On the supplied information, the principal risk is uncertainty: a potentially material change to data-use or AI-training rights cannot be evaluated from the word-count description alone.

Summary of Important Changes and Risks

1. Customer Data and AI Model Training — Major Change

• The revised terms expressly state that Vercel may use Your Content to train artificial intelligence and machine-learning models.
• They also permit Vercel to share Your Content with third parties, including third-party product providers, for training and improving their AI and machine-learning models (“Model Training”).
• The diff indicates that Model Training is not available or is treated differently for customers on a paid Pro plan, but the excerpt does not include the complete sentence or explain the applicable opt-out or exclusion. This should be confirmed in the final text.
• The customer grants Vercel a broad, worldwide, royalty-free, sublicensable and transferable license to use, copy, modify, reproduce, distribute, display, publish, store, perform and create derivatives of Your Content. The stated purposes now expressly include:
• Providing and improving the Services;
• Developing new products and services;
• Securing and protecting the Services and third parties from fraud, abuse, malware and malicious content; and
• AI and machine-learning model training and improvement.
• Risk: Customer code, prompts, data, text, confidential information or personal data submitted to the Services may be incorporated into training processes or disclosed to third-party providers, potentially creating confidentiality, intellectual-property, privacy, regulatory and trade-secret concerns. Customers should verify whether business, personal, health or regulated data is excluded, de-identified, deleted, or subject to an opt-out.

2. Broader Third-Party AI Functionality

• The terms now more clearly address Vercel AI features, customer-controlled AI tools, AI agents and third-party tools.
• Customers are responsible for actions taken by AI functionality or third-party tools acting as their “electronic agent,” including resulting costs and use of the Services.
• Risk: Customers may be liable for autonomous AI actions, unauthorized deployments, excessive usage, policy violations or third-party charges.

3. Dispute Resolution and Class Waiver

• Arbitration and class-action waiver provisions are prominently incorporated into the acceptance language and expressly stated to be mutual.
• The agreement requires individual, binding arbitration, generally under California law and JAMS procedures, while waiving class, collective and representative proceedings.
• A 30-day opt-out process is specified.
• Risk: Customers may lose access to court and class proceedings and may face procedural and cost burdens in pursuing claims.

4. Customer and Account Obligations

• The revised structure emphasizes that users must be at least 16, provide accurate account information, maintain confidentiality, use unique credentials and accept responsibility for all account activity.
• Vercel retains the right to access accounts for technical support.
• Risk: Customers bear broader operational responsibility for unauthorized access and activity under their accounts.

5. Other Notable Clarifications

• Open-source software, including Next.js, is expressly excluded from the Agreement and governed by its own licenses.
• Data processing is tied to Vercel’s DPA and Privacy Policy, with Vercel acting as controller or processor depending on the data and context.
• Vercel’s rights to use customer feedback, names, trademarks and logos remain broad.

Summary

The diff states only that approximately 76 words were removed, but does not identify which words, clauses, or sections were deleted.

AI Training and Data Use

• No determination is possible regarding whether the deletion changes how the customer’s data may be used to train, fine-tune, evaluate, or improve AI models.
• The removed language could potentially have addressed:
• Whether customer data may be used for AI training;
• Whether data is used only to provide the services;
• Whether customer data is anonymized, aggregated, or de-identified first;
• Whether human reviewers or service providers may access the data;
• Whether the customer may opt out of AI training;
• Ownership of inputs, outputs, or model improvements; or
• Retention and deletion of data used in model development.

Legal and Commercial Risk

• Material uncertainty: A deletion may remove an express customer protection, restriction, consent requirement, or disclosure. Alternatively, it may remove a provider-favorable permission. The direction of the risk cannot be determined without the actual text.
• Potential implied rights: If restrictions on AI training were deleted, the remaining agreement may give the provider broader discretion to use customer data, subject to any surviving privacy, confidentiality, or data-processing terms.
• Potential loss of safeguards: Deleting limitations, opt-out rights, deletion obligations, or purpose restrictions could materially increase the customer’s privacy, confidentiality, regulatory, and intellectual-property risk.
• Interpretation risk: The remaining language should be reviewed for conflicts or gaps concerning customer data ownership, permitted uses, confidentiality, security, retention, subprocessors, and model training.

Recommended Follow-Up

Obtain a redline showing the actual deleted language and compare it with the surviving provisions, especially clauses titled:

• Data Use or Customer Data;
• Artificial Intelligence or Machine Learning;
• Confidentiality;
• Privacy/Data Processing;
• Service Improvements;
• Training, Evaluation, or Model Development; and
• Data Retention and Deletion.

Without the deleted text and surrounding provisions, no reliable conclusion can be reached about the legal effect of the change or whether customer data may now be used to train AI models.

Key Changes and Risks

1. Customer Content and AI Model Training

• The terms continue to grant Vercel a worldwide, royalty-free, sublicensable, transferable license to use, copy, modify, reproduce, distribute, display, store, perform, and create derivative works from “Your Content.”
• The permitted purposes expressly include:
• Providing and improving the Services;
• Developing new products and services;
• Securing and protecting the Services and third parties from fraud, abuse, malware, malicious files, and similar threats.
• The terms also state that Vercel may use Your Content to train artificial-intelligence and machine-learning models and may share Your Content with third-party product providers, including for training and improving their models.
• The diff does not clearly show a new opt-out right or a narrower definition of training. It indicates that paid Pro-plan customers are excluded from Model Training, but the excerpt does not show the full scope, exceptions, or whether other paid plans receive the same protection.
• Risk: Unless the customer is covered by the Pro-plan exclusion or another written agreement, submitted code, data, prompts, and other materials may be used for AI training and shared with third parties. The license is broad enough to permit modification and derivative use, and it does not appear limited to anonymized or aggregated data.

2. Arbitration and Class-Action Waiver

• The mutual arbitration clause and class-action waiver are now prominently emphasized near the beginning of the agreement.
• Customers agree to resolve most disputes through final and binding arbitration and waive participation in class, collective, multiple-plaintiff, and representative actions.
• A 30-day opt-out process is specified, including a required email address and subject line.
• Risk: Customers may lose access to court proceedings and class remedies unless they timely opt out. Arbitration, mass-arbitration procedures, San Francisco venue provisions, and potential fee exposure may increase enforcement costs.

3. Account and Eligibility Obligations

• The terms clarify that users must be at least 16 years old, provide accurate signup information, update it regularly, and maintain confidential credentials.
• Each user must have unique login credentials and may not share them.
• Customers remain responsible for all activity under their accounts.
• Risk: Organizations may face liability for unauthorized activity, credential sharing, or inaccurate account information.

4. Third-Party and AI Tools

• AI functionality and third-party tools are expressly addressed.
• Tools acting on the customer’s behalf are treated as an “electronic agent” under the Uniform Electronic Transactions Act.
• Customers are responsible for actions and costs incurred by those tools.
• Risk: Automated actions may legally bind the customer, with limited ability to disclaim unauthorized or unintended activity.

5. Other Notable Changes

• Additional incorporated terms now cover AI Products, APIs, DORA, Marketplace services, domain registration, and non-Vercel services.
• Vercel retains broad rights to access accounts for technical support.
• The agreement continues to permit unilateral updates by posting or sending notice, creating an ongoing obligation to monitor revisions.

Summary of Important Changes and Risks

1. Customer Content may be used for AI model training

The most significant change is the express addition of a “Model Training” provision. It states that:

• Vercel may use Your Content to train Vercel’s artificial intelligence and machine-learning models.
• Vercel may share Your Content with third-party products, including for training and improving their AI and machine-learning models.
• The customer grants Vercel a broad license to use, copy, modify, reproduce, distribute, store, perform, and create derivative works from Your Content.
• The license is worldwide, royalty-free, sublicensable, transferable, and appears to continue for purposes including improving Services, developing products and services, security, and fraud or abuse prevention.

Risks

• No clear limitation to de-identified, aggregated, or anonymized data appears in the diff.
• The provision may cover source code, proprietary software, confidential business information, personal data, and other materials uploaded to Vercel.
• Sharing with third parties creates additional confidentiality, intellectual-property, privacy, and security risks.
• The language does not clearly provide a general opt-out right. It states that if a customer is on a paid Pro plan, “Model Training is not …,” but the remainder is omitted from the diff. The exact scope—such as whether training is prohibited, opt-in, or subject to exceptions—should be confirmed.
• Customers should verify that they have authority and appropriate consents to permit this use, especially for employee data, personal information, regulated data, customer data, and third-party code.

2. New AI functionality and autonomous tools

The revised Terms add a dedicated section covering AI functionality and third-party tools. Customers may authorize Vercel or third-party AI tools, agents, or similar technologies to access Services.

Risks

• AI systems capable of autonomous action may make changes or incur costs on the customer’s behalf.
• The customer is responsible for actions taken by third-party tools and for costs resulting from their use.
• Third-party providers may have separate data-use, retention, and training practices.

3. Broader and reorganized content license

The prior technical-support access language is replaced or incorporated into a substantially broader license covering product improvement, new products and services, security, and protection against fraud, abuse, malware, and malicious content.

This increases Vercel’s stated rights to process and derive value from customer-submitted materials.

4. Arbitration and class-action waiver made prominent

The revised introduction prominently requires review and acceptance of a mutual arbitration clause and class-action waiver. It also adds or clarifies mass-arbitration procedures and a 30-day opt-out process.

Risk

Customers may lose the ability to bring court claims or participate in class, representative, or multi-plaintiff proceedings unless they timely opt out.

5. Other notable changes

• Open-source software, including Next.js, is expressly excluded from the Terms and governed by its licenses.
• Account requirements are more explicit: users must be at least 16, provide accurate information, and use unique, non-shared credentials.
• AI Products and Services, Marketplace offerings, APIs, DORA, and other services are incorporated through additional terms, increasing the importance of reviewing linked documents.

Summary

Scope of the Change

The diff states only that approximately 76 words were removed from the document. It does not identify:

• Which clauses were deleted;
• Whether the deletions were complete sentences, definitions, or substantive provisions;
• Any replacement language; or
• Whether the deleted text concerned data use, artificial intelligence, confidentiality, security, or liability.

AI Training and Customer Data

No specific change regarding the use of customer data to train AI models can be identified from the supplied diff.

The deletion could be significant if the removed language addressed any of the following:

• Permission to use customer data, content, prompts, outputs, or metadata to train, fine-tune, or improve AI models;
• Whether customer data would be used for service improvement or product development;
• Opt-out or consent requirements;
• Use of data in aggregated, de-identified, or anonymized form;
• Human review or access to customer data;
• Retention or deletion of data used for training;
• Ownership of model improvements or AI outputs; or
• Restrictions on using confidential, personal, regulated, or sensitive information.

If any such language was removed, the practical effect could be either more protective or less protective, depending on what was deleted. For example, removing a training authorization may restrict the provider’s rights, while removing an opt-out, confidentiality safeguard, or deletion obligation may increase the customer’s risk.

Other Potential Legal Risks

Because the actual deleted text is not provided, it is not possible to determine whether the change affects:

• Confidentiality obligations;
• Data-processing or privacy compliance;
• Security commitments;
• Intellectual-property ownership or licensing;
• Warranties and disclaimers;
• Indemnities;
• Limitation of liability;
• Termination or data-return obligations; or
• Regulatory requirements.

Recommended Next Step

Obtain the actual 76-word deletion and any surrounding language—preferably the full before-and-after clause. The deleted text should be reviewed specifically for terms such as “train,” “training,” “improve,” “machine learning,” “artificial intelligence,” “models,” “customer data,” “content,” “prompts,” “outputs,” “de-identified,” “aggregated,” “retain,” and “opt out.” Without that text, no reliable conclusion can be reached about the legal effect of the amendment.

Summary

The provided diff only states that approximately 76 words were removed from the document. It does not identify the deleted language or show any surrounding provisions.

AI Training and Customer Data

• It is not possible to determine whether the changes affect:
• The provider’s right to use customer data to train, fine-tune, or improve AI models;
• Use of customer prompts, inputs, outputs, or usage metadata;
• Whether data may be used in aggregated, de-identified, or identifiable form;
• Opt-out or consent requirements;
• Data retention, deletion, or isolation commitments;
• Restrictions on using customer data to train models made available to other customers; or
• Confidentiality, security, or intellectual-property protections relating to AI training.

Potential Legal Significance

Deletion of language could:

• Remove a customer protection or limitation on the provider’s data-use rights;
• Eliminate an express prohibition on AI training;
• Narrow confidentiality, deletion, or security obligations;
• Remove an opt-out mechanism or consent requirement; or
• Conversely, eliminate a provider permission, potentially restricting service improvement or model training.

The legal effect depends entirely on the exact words removed and the remaining contract language.

Assessment Limitation

No reliable risk assessment can be made from a word-count description alone. The actual deleted text, and preferably the surrounding unchanged provisions, is required to determine whether the customer’s data may be used to train AI models and whether the change increases or decreases customer risk.

Structured Summary of Important Changes

1. Customer Content may be used for AI model training — major change

• The revised language expressly states that Vercel may use “Your Content” to train artificial intelligence and machine-learning models.
• It also permits Vercel to share Your Content with third parties, including for those parties’ training and improvement of their AI and machine-learning models.
• “Your Content” is broadly defined to include software, source code, data, text, feedback, suggestions, inputs, and other materials submitted, stored, or transmitted through the Services.
• The license granted to Vercel is worldwide, royalty-free, fully paid, sublicensable, transferable, and includes rights to use, copy, modify, reproduce, distribute, perform, and create derivative works.
• The stated purposes now include providing and improving the Services, developing new products and services, security, and protecting Vercel and third parties from fraud, abuse, malware, and malicious content.
• The diff appears to include a limitation or exclusion concerning paid Pro plans (“If you are on a paid Pro plan, Model Training is not…”), but the excerpt is incomplete. The final text should be reviewed to determine whether Pro customers are automatically excluded, may opt out, or receive some other protection.
• The language does not appear to provide a general customer opt-out, deletion mechanism, anonymization requirement, restriction on training with confidential information, or prohibition on third-party model training. Customers should assume that sensitive, proprietary, personal, or regulated data may be used or disclosed unless the final terms or a separate plan/DPA clearly provide otherwise.

2. Broader operational license to Vercel

The revised terms appear to replace a narrower right to access customer accounts for technical support with a broader license allowing Vercel to use customer content for product development, service improvement, security, and fraud prevention. This increases Vercel’s permitted uses and may affect confidentiality, intellectual-property, and data-governance expectations.

3. AI and third-party tools

• New or expanded provisions address AI functionality, AI providers, autonomous actions, and third-party tools.
• Customer-authorized AI tools are treated as the customer’s “electronic agent” under UETA or equivalent law.
• Customers bear responsibility for actions taken by those tools and for costs incurred through their use of the Services.

4. Dispute-resolution changes

The revised terms prominently add or reorganize a mandatory individual arbitration clause and class-action waiver, governed by California law and generally requiring proceedings through JAMS. A 30-day opt-out procedure is provided, but failure to opt out may substantially limit court and class-action rights.

5. Other notable changes

• Eligibility is stated as requiring users to be at least 16.
• Each user must have unique, non-shared credentials; customers remain responsible for account activity.
• Additional incorporated terms now cover AI Products, APIs, DORA, domains, Marketplace services, and non-Vercel services.
• Vercel receives a license to use customer names, logos, and trademarks in marketing materials unless separately restricted.

*Because the supplied diff is fragmented and incomplete, the final published wording—especially the Pro-plan Model Training exception—should be confirmed before relying on this summary.*

Key Changes and Risks

1. Customer Content and AI Model Training

• The revised terms expressly state that Vercel may use Your Content to train artificial-intelligence and machine-learning models.
• They also permit Vercel to share Your Content with third-party providers, including for training and improving those providers’ AI and machine-learning models.
• “Your Content” is defined broadly to include software, code, data, information, feedback, text, inputs, and other materials submitted to or stored through the Services.
• The license granted to Vercel is worldwide, royalty-free, fully paid, sublicensable, transferable, and includes rights to use, copy, modify, reproduce, distribute, display, publish, store, perform, and create derivative works.
• The stated purposes now include:
• Providing and improving the Services;
• Developing new products and services;
• Securing and protecting the Services; and
• Protecting Vercel and third parties from fraud, abuse, malware, malicious files, viruses, and similar threats.

AI-related risks

• Confidentiality and ownership risk: Customer code, proprietary information, personal data, or regulated data could be incorporated into model-development processes or disclosed to third-party AI providers.
• Broad downstream use: “Sublicensable” and “transferable” rights may allow contractors, affiliates, and other providers to receive or use the content.
• Limited control over outputs: The terms do not appear to promise that customer content will be deleted from training datasets or that trained models can be “unlearned.”
• Opt-out limitation: The diff indicates that paid Pro-plan customers are treated differently, stating that Model Training is not [permitted or applicable— the provided excerpt is incomplete]. The exact scope, default setting, and availability of any opt-out should be confirmed in the complete language and applicable plan terms. No comparable protection is apparent for free, hobby, or other plans.
• Customers remain responsible for ensuring they have the rights and permissions necessary for Vercel’s use and sharing of the content.

2. New AI Functionality and Third-Party Tools

• A new section covers AI features, AI providers, third-party models, agents, and tools.
• AI functionality may take autonomous actions, potentially affecting projects or deployments.
• Customer-enabled AI or third-party tools are deemed the customer’s “electronic agent” under UETA or equivalent law.
• Customers bear responsibility for actions taken and costs incurred by those tools.

3. Dispute Resolution

• The revised agreement prominently adds mutual binding arbitration and a class-action waiver, incorporated into the main terms and Schedule I.
• Claims generally must be brought individually, with limited exceptions.
• Customers have a 30-day opt-out process, but the excerpt states that if a customer opts out, Vercel also will not be bound by the arbitration provisions.

4. Other Material Changes

• The terms are substantially reorganized and incorporate numerous additional policies and addenda, including AI Product Terms, DORA, API, Marketplace, domain-name, and non-Vercel-services terms.
• Vercel may access accounts for technical support.
• Customers have heightened responsibility for account activity, unique credentials, content rights, security configuration, and legal compliance.
• Vercel retains broad rights in system data and may change service limits and terms over time.

Summary

The diff indicates that approximately 76 words were removed from the document, but it does not identify which words, clauses, or sections were deleted.

AI Training and Customer Data

• No reliable conclusion can be drawn about changes to the use of customer data for training AI models.
• The deleted language may have addressed:
• Whether customer data may be used to train, fine-tune, or improve AI models;
• Whether data is used in aggregated, de-identified, or identifiable form;
• Whether customer prompts, inputs, outputs, or metadata are retained;
• Whether the provider may share data with affiliates or third-party AI providers;
• Opt-out or consent rights; or
• Security, confidentiality, and deletion obligations.
• If the removed language limited AI training or required customer consent, its deletion could expand the provider’s rights to use customer data.
• If the removed language authorized AI training, its deletion could instead restrict those rights or create ambiguity.

Other Potential Legal Risks

Because the specific deletions are not provided, it is also impossible to determine whether the changes affect:

• Confidentiality and privacy protections;
• Data ownership or licensing rights;
• Liability, indemnification, or limitations of liability;
• Data retention and deletion;
• Security commitments;
• Compliance obligations; or
• Termination and post-termination data handling.

Recommended Next Step

Obtain a redline showing the exact 76 deleted words, preferably with the surrounding provisions and section headings. The impact of the amendment cannot be assessed accurately from the deletion count alone.

Key Changes and Risks

1. Customer Content may be used for AI training

The revised Terms expressly state that Vercel may use “Your Content” to train artificial intelligence and machine-learning models. This is a material expansion or clarification of Vercel’s data-use rights.

The license to Your Content remains broad and includes rights to:

• Use, copy, modify, adapt, reproduce, distribute, display, publish, store, perform, and create derivative works;
• Provide and improve the Services;
• Develop new products and services;
• Secure and protect the Services and third parties from fraud, abuse, malware, and malicious content; and
• Train and improve AI and machine-learning models.

Vercel may also share Your Content with third-party providers, including for training or improving their AI and machine-learning models.

Customer risks

• Confidential source code, prompts, business information, personal data, or proprietary materials submitted to the Services may be incorporated into model-development activities.
• The license is worldwide, royalty-free, transferable, and sublicensable, with no express deletion, anonymization, retention, or model-unlearning commitment in the quoted language.
• “Your Content” is defined broadly and may include data stored or transmitted through projects and deployments.
• Customers should verify whether the full Model Training section provides an opt-out, consent requirement, or plan-based restriction. The diff states that Model Training is not [permitted or applicable] for paid Pro plans, but the remainder of that sentence is not included, so the exact protection cannot be confirmed from this diff.

2. Broader and clearer AI functionality provisions

A new or expanded section addresses:

• Vercel AI products and features;
• Access to third-party AI providers and their models;
• AI functionality capable of taking autonomous actions; and
• Customer-enabled AI agents and third-party tools.

Customers are responsible for actions taken through their AI functionality or third-party tools. Such tools are treated as the customer’s electronic agent, potentially increasing the customer’s responsibility for unauthorized, erroneous, or costly actions.

3. Account and security obligations expanded

The revised text adds or restores requirements that:

• Each user must have unique credentials;
• Credentials cannot be shared;
• Customers are responsible for all activities under their account; and
• Account information must be accurate and regularly updated.

This may increase liability for employee, contractor, or AI-agent activity.

4. Dispute terms made substantially more prominent

The revised Terms prominently require review and acceptance of mutual binding arbitration and a class-action waiver. The provisions require individual arbitration, generally in or near San Francisco, and waive participation in class, representative, and multiple-plaintiff proceedings, subject to stated exceptions and a 30-day opt-out process.

5. Other notable changes

• Open-source software, including Next.js, is expressly excluded from the Agreement and governed by its own licenses.
• AI Product Terms, DORA terms, API terms, and other addenda are incorporated by reference.
• Vercel retains broad rights to access accounts for technical support.
• Vercel may use customer names, trademarks, and logos in marketing materials.
• Pricing, payment, termination, disclaimer, and liability provisions remain strongly Vercel-favorable.

Summary

The diff states only that approximately 76 words were removed from the document. It does not identify which words, clauses, or sections were deleted.

Legal and Commercial Impact

• The specific legal effect of the deletions cannot be determined without the redlined text or the original and revised versions.
• Deleted language could affect important provisions such as:
• Data ownership and licensing rights
• Confidentiality and permitted disclosures
• Use of customer data for analytics, product improvement, or AI training
• Service-provider and subprocessors’ rights
• Security and privacy obligations
• Liability, indemnification, or limitations of liability
• Data retention and deletion
• Audit, compliance, or termination rights

AI-Training Provisions

No conclusion can be reached about changes to AI-model training based solely on the statement that 76 words were removed.

In particular, the diff does not reveal whether the deleted language:

• Authorized or prohibited using customer data to train, fine-tune, or improve AI models;
• Limited AI training to aggregated, de-identified, or anonymized data;
• Required customer consent or provided an opt-out;
• Allowed use of customer prompts, inputs, outputs, personal data, or confidential information;
• Granted the provider a license to use customer content;
• Restricted human review or model-training data retention; or
• Applied different rules to public, business, or enterprise customers.

Risk Assessment

The principal risk is lack of visibility into the deleted terms. If the removed language concerned data use or AI training, its deletion could either:

1. Remove a customer protection, potentially broadening the provider’s rights through remaining general-purpose language; or

2. Remove a provider authorization, potentially restricting data use or creating ambiguity.

The complete redline, or at least the 76 deleted words and their surrounding provisions, is required for a reliable legal analysis.

Summary

The diff only states that approximately 76 words were removed; it does not identify which words, clauses, or sections were deleted.

AI Training and Data Use

• It is not possible to determine whether the deleted language affected:
• The customer’s data being used to train, fine-tune, or improve AI models;
• Restrictions or permissions concerning use of customer content;
• De-identification, aggregation, or anonymization requirements;
• Whether inputs, outputs, prompts, or usage data may be retained;
• Opt-out rights or consent requirements;
• Ownership of customer data or AI-generated outputs; or
• Confidentiality and security protections applicable to data used for model development.

Potential Legal Risks

• If the removed text previously restricted AI training or secondary use, its deletion could materially expand the provider’s rights to use customer data.
• If the removed text contained consent, notice, opt-out, deletion, or confidentiality protections, customers may have fewer contractual safeguards.
• Conversely, the deletion could also remove language that was unfavorable to the customer, such as broad warranties, indemnities, or compliance obligations.
• Because the actual deleted language is unavailable, the direction and significance of the change cannot be reliably assessed.

Required Information

Please provide the actual 76-word deletion, preferably with the surrounding clause or a tracked-changes version. Without the deleted text, no definitive conclusion can be reached regarding changes to AI-training rights or other legal obligations.

Key Changes and Risks

1. Customer Content may be used for AI model training

• The revised terms expressly state that Vercel may use “Your Content” to train artificial intelligence and machine-learning models.
• The license granted to Vercel is broad: worldwide, royalty-free, fully paid, sublicensable, transferable, and includes rights to use, copy, modify, adapt, reproduce, distribute, display, publish, store, perform, and create derivative works.
• The stated purposes now include:
• Providing and improving the Services;
• Developing new products and services;
• Securing and protecting the Services;
• Protecting Vercel and third parties from fraud, abuse, malware, malicious files, viruses, and similar risks; and
• Training and improving Vercel’s and third-party AI and machine-learning models.
• Vercel may share Your Content with third-party products or providers, including for their own AI-model training and improvement.
• The excerpt states: “If you are on a paid Pro plan, Model Training is not…” The sentence is incomplete in the diff. The full terms should be reviewed to determine whether Pro customers are excluded automatically, must opt in, or have an opt-out right.
• Risk: Confidential source code, personal data, customer information, or regulated data submitted to the Services may be incorporated into training processes or shared with third parties unless a contractual exclusion, opt-out, or technical restriction applies. The terms do not show clear limits on retention, anonymization, model deletion, or whether trained models can later be used commercially.

2. Broader content license

The former support-focused right to access accounts has been replaced or supplemented by a much broader license to use and exploit Customer Content. The license is transferable and sublicensable, potentially allowing affiliates, contractors, and third-party AI providers to process the data.

Customers remain responsible for ensuring they own the content and that its use does not infringe privacy, publicity, intellectual-property, or other rights.

3. New or expanded AI functionality provisions

• New provisions address Vercel AI features, third-party AI providers, autonomous actions, AI agents, and third-party tools.
• Customers are responsible for actions taken by AI functionality or tools acting as their “electronic agent,” including associated costs.
• Risk: Customers may bear responsibility for automated actions, errors, unauthorized changes, or charges generated by AI tools connected to their account.

4. Dispute-resolution changes and emphasis

• The agreement now prominently highlights binding arbitration and class-action waiver requirements.
• Claims generally must be brought individually, not through class, representative, or multiple-plaintiff proceedings.
• A 30-day opt-out process is specified, with detailed JAMS procedures and mass-arbitration provisions.
• California law and San Francisco-related arbitration/forum provisions remain important.

5. Other notable changes

• The terms expressly require users to be at least 16 and maintain accurate account information.
• Open-source software, including Next.js, is expressly excluded from the main Agreement and governed by its licenses.
• Data-processing terms are tied to the DPA and Privacy Policy; Vercel may act as controller or processor.
• Vercel retains ownership of System Data, excluding Account Information.
• Customers grant Vercel rights to use their names, trademarks, and logos in marketing materials.

Summary of Important Changes

1. Customer Content may be used for AI model training

• The revised terms expressly state that Vercel may use Your Content to train artificial intelligence and machine-learning models.
• They also permit Vercel to share Your Content with third-party providers, including for those providers’ training and improvement of AI and machine-learning models.
• The license granted to Vercel remains broad: worldwide, royalty-free, fully paid, sublicensable, transferable, and covering copying, modification, reproduction, distribution, display, storage, performance, and creation of derivative works.
• The stated purposes now expressly include:
• Providing and improving the Services;
• Developing new products and services;
• Securing and protecting the Services; and
• Protecting Vercel and third parties from fraud, abuse, malware, malicious files, and similar threats.
• The revised language appears to add a plan-specific limitation stating that Model Training is not available—or may be restricted—if the customer is on a paid Pro plan. The excerpt is incomplete, so the exact scope, opt-out process, and whether the restriction applies to all content or only certain uses should be confirmed in the full terms.

Risk: Customers may have limited control over whether submitted code, prompts, data, or other materials are used to train models or disclosed to AI providers. Confidential, proprietary, personal, regulated, or third-party data should not be submitted unless the customer’s permissions, contracts, and compliance obligations allow this use.

2. New AI and third-party-tool provisions

• A new section addresses AI Functionality and third-party tools, agents, and similar technologies.
• If enabled, these tools may take autonomous actions through the Services.
• Customers are responsible for actions taken on their behalf and for costs incurred through third-party tools.
• Third-party tools are characterized as the customer’s “electronic agent” under UETA or equivalent law.

Risk: Customers may bear responsibility for automated actions, unauthorized changes, consumption charges, and errors caused by connected AI agents.

3. Arbitration and class-action waiver made prominent

• The revised introduction prominently requires review and acceptance of a mutual binding arbitration clause and class-action waiver.
• Claims generally must be brought individually, not as class, representative, or multiple-plaintiff actions.
• California law, San Francisco venue, JAMS procedures, mass-arbitration procedures, and a 30-day opt-out process are specified.

4. Account and security responsibility clarified

• Each user must have unique credentials; credential sharing is prohibited.
• Customers are responsible for all activities occurring under their accounts, increasing the practical consequences of compromised credentials or misuse by users and connected tools.

5. Additional incorporated terms

• AI Products and Services are subject to separate AI Product Terms.
• Marketplace, domain registration, non-Vercel services, DORA, API, and other addenda are incorporated, potentially adding obligations and limitations.

Summary

The diff indicates that approximately 76 words were removed, but it does not identify which provisions were deleted.

AI Training and Data Use

• It is not possible to determine from the provided information whether the deletions:
• Permit or restrict using customer data to train, fine-tune, or improve AI models;
• Change whether customer data may be used for product development, analytics, or service improvement;
• Remove consent, notice, opt-out, or approval requirements;
• Change ownership or licensing rights in customer data, inputs, outputs, or derived data; or
• Affect requirements to anonymize, de-identify, aggregate, or delete customer data before AI training.

Potential Legal Risks

Because the deleted language is unavailable, the principal risk is uncertainty. Removing approximately 76 words could materially:

• Broaden the provider’s rights to use customer content or usage data;
• Eliminate restrictions on retaining or reusing customer data;
• Remove confidentiality, security, deletion, or purpose-limitation protections;
• Alter compliance obligations relating to privacy, data protection, or regulated information; or
• Remove customer consent or contractual remedies.

Alternatively, the deletions could have reduced the provider’s rights or added customer protections. The direction of the change cannot be reliably determined from the description alone.

Required Information

To perform a meaningful legal comparison, the actual deleted text and any replacement or surrounding provisions are required. In particular, provide the sections addressing:

• Customer data and content;
• AI or machine-learning model training;
• Product or service improvement;
• Data retention and deletion;
• Confidentiality and security; and
• Ownership, licenses, and derived data.

Conclusion: No specific change to AI-training rights or other legal obligations can be confirmed from the information supplied.

Important Changes and Risks

1. Customer Content may be used for AI model training

The revised terms add an express right for Vercel to:

• Use Your Content to train Vercel’s artificial intelligence and machine-learning models.
• Share Your Content with third parties, including for training and improving those parties’ AI and machine-learning models.
• Use Your Content to provide and improve Services, develop new products and services, secure and protect the Services, and address fraud, abuse, malware, malicious files, and similar threats.

Key risks

• This is a significant expansion beyond ordinary service delivery and technical-support purposes.
• The license is worldwide, royalty-free, sublicensable, transferable, perpetual in effect for permitted uses, and broad enough to include copying, modification, reproduction, distribution, performance, and creation of derivative works.
• The clause does not appear to limit AI training to anonymized, aggregated, or de-identified data.
• There is no clear restriction on the types of third parties receiving the data, the geographic locations involved, retention periods, or whether trained models can retain or reproduce information from Customer Content.
• Customers may need to ensure they have obtained all required consents and rights before submitting personal data, confidential information, code, or third-party materials.

The diff states: “we may use Your Content to train our artificial intelligence and machine learning models” and may share it with third parties for their model training and improvement.

The text also indicates that Model Training may not apply to paid Pro plans, but the provided diff does not show the complete limitation or any applicable opt-out process. Customers should confirm:

• Which plans are excluded;
• Whether the exclusion applies automatically;
• Whether enterprise or team settings control training;
• Whether previously submitted data is removed from training pipelines; and
• Whether third-party AI providers have equivalent restrictions.

2. AI functionality and third-party tools are expressly addressed

A new or expanded section covers AI Functionality, AI providers, agents, and third-party tools. Customers are responsible for:

• Costs incurred through third-party tools;
• Actions taken by AI Functionality or third-party tools on the customer’s behalf; and
• The tools’ access to the Services.

Third-party tools are characterized as an “electronic agent” under UETA or an equivalent legal concept. This may make customers responsible for automated actions, including unauthorized, erroneous, or harmful actions taken by connected AI systems.

AI Products and Services are also made subject to separate incorporated AI Product Terms, creating additional obligations and potentially allowing terms to be updated separately.

3. Dispute rights are materially emphasized

The revised presentation prominently requires review and acceptance of a mutual arbitration clause and class-action waiver. The agreement adds or clarifies:

• Final and binding arbitration;
• Individual—not class, collective, or representative—claims;
• California law and generally San Francisco venue;
• A 30-day arbitration opt-out procedure; and
• Special mass-arbitration procedures.

This may substantially limit the customer’s ability to sue in court or participate in a class action.

4. Other notable changes

• The terms now more clearly cover Vercel affiliates and a broader range of Services.
• Open-source software, including Next.js, is expressly excluded from the main Agreement and governed by its applicable licenses.
• Customers remain responsible for all activities occurring under their accounts, with stronger unique-login and credential-sharing requirements.
• Vercel retains ownership of System Data and may use it during and after the Agreement, subject to the stated definition and exclusions.