accessiBe

Summary

The provided diff states only: “Removed approximately 3 words from the document.”

Key Findings

• The specific deleted words are not identified.
• No additions, replacements, or substantive clause text are shown.
• It is therefore impossible to determine whether the change affects:
• Customer data rights or ownership
• Permission to use customer data
• Confidentiality or security obligations
• Data retention or deletion
• Liability, indemnification, or compliance obligations
• Use of customer data to train, fine-tune, evaluate, or improve AI models

AI Training/Data-Use Risk

No conclusion can be reached about changes to AI-model training. The deleted language could potentially narrow or expand the provider’s rights to use customer data, but the actual impact cannot be assessed without the missing text.

Recommended Follow-Up

Provide the exact three deleted words and the surrounding sentence or clause, preferably using the original diff notation. This is necessary to determine whether the deletion changes the customer’s consent, the provider’s data-use rights, or any restrictions on AI training.

Summary of Important Changes

1. AI-related processing of customer data

Call recording and AI processing added

The notice now states that accessiBe may:

• Record calls for customer support, quality assurance, training, and marketing.
• Process recordings using AI tools to create transcripts, summaries, and notes.
• Inform callers at the start of the call when recording occurs.
• Provide an opportunity to opt out of recording.

Risks and implications:

• Call content may contain confidential business information, personal data, or sensitive information supplied by the customer.
• The notice does not clearly identify the AI providers, whether recordings or transcripts are transferred to third parties, how long they are retained, or whether they are used to train general-purpose AI models.
• It does not expressly state that customer call data will not be used to train AI models. The reference to “training” appears to concern employee or service-quality training, but the language could be interpreted broadly.
• Customers may need to provide their own call-recording notices and obtain legally required consent from participants, particularly in jurisdictions requiring all-party consent.

No express authorization to train models identified

The diff does not appear to add an express provision authorizing accessiBe to use customer data, website content, end-user data, support communications, or call recordings to train AI models.

However, the new AI-processing language is incomplete from a risk perspective because it does not expressly distinguish between:

1. Using AI tools to process data and generate outputs; and

2. Retaining or using that data to train, improve, or fine-tune AI models.

Customers should seek confirmation that their data will not be used for model training unless separately authorized.

2. End-user data and customer responsibility

The notice states that accessiBe may process end-user IP addresses, URLs, accessibility preferences, and other information while users operate accessWidget. The customer is responsible for making commercially reasonable efforts to disclose this processing and obtain consent.

AccessiBe characterizes itself as a processor of customer user data, with processing governed by the Data Processing Addendum.

Risk: The customer bears an explicit compliance burden for notices and consent involving its website users. The scope of “other information that is not Personal Information” is also broad and may require further clarification.

3. Marketing use of data from one-time services

For limited or one-time services, such as accessScan, the customer agrees that accessiBe may use submitted contact information for marketing communications for up to six months.

The notice characterizes this as contractual consideration under GDPR Article 6(1)(b).

Risk: Treating marketing permission as part of the service contract may be challenged in some jurisdictions, where consent or a separate legitimate-interest analysis may be required. The notice does preserve a right to object.

4. Other notable additions or clarifications

• Support chatbot inputs may be processed in full, not merely the email address.
• Calls may be recorded for marketing purposes.
• Broader descriptions of analytics, cookies, affiliated-company sharing, international transfers, retention, and California privacy rights are included.
• The notice permits retention of aggregated anonymous data indefinitely.

Summary of Available Changes

Scope of the Diff

The diff states only that approximately 5,964 words were removed from the document. It does not identify:

• Which provisions were deleted;
• Whether any provisions were added or replaced;
• The subject matter of the deleted language; or
• Whether the deleted language concerned data use, confidentiality, intellectual property, AI, security, or other topics.

AI Model Training and Customer Data

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

In particular, the diff does not show whether the agreement now:

• Permits or prohibits using customer data to train, fine-tune, or improve AI models;
• Allows use of customer data in aggregated, de-identified, or anonymized form;
• Restricts human review or labeling of customer data;
• Allows retention of customer data for model-development purposes;
• Gives the provider ownership or broad usage rights over inputs, outputs, or derived data;
• Requires consent before customer data is used for AI training; or
• Requires deletion or segregation of customer data from training datasets.

Potential Legal Risks From the Deletion

Although the substance cannot be assessed, removing approximately 5,964 words could create material risk if the deleted provisions included protections such as:

• Confidentiality: Reduced restrictions on disclosure or use of customer information.
• Data-use limitations: Removal of purpose restrictions could expand permitted uses of customer data.
• AI safeguards: Loss of express prohibitions or consent requirements for model training.
• Security obligations: Fewer security controls, incident-notification duties, or audit rights.
• Data ownership: Weaker protections regarding ownership of customer data, inputs, outputs, or derived information.
• Retention and deletion: Less certainty about how long data may be retained or whether it will be deleted.
• Indemnification and liability: Reduced remedies or liability allocation for misuse, breaches, or unauthorized data use.
• Regulatory compliance: Removal of obligations relevant to privacy and data-protection laws.

Conclusion

The supplied diff is insufficient to perform a substantive clause-by-clause analysis. The only confirmed change is a large-scale deletion. The complete redlined text, or at least the deleted and surrounding provisions, is needed to determine whether customer data may now be used for AI training and whether important contractual protections have been removed.

Summary of Changes

The provided diff contains only the following apparent text:

> “accessiBe Terms of Service | Legal Agreement | accessiBe [Skip to content] Connecting you with the right rep...”

No additions, deletions, or replacements are identifiable using the specified notation:

• Additions: {...}
• Deletions: [...]
• Replacements: [...] {...}

Legal Significance

Because the substantive contract language is not included, it is not possible to determine whether the Terms of Service changed in areas such as:

• Customer data ownership or licensing
• AccessiBe’s rights to process, store, or disclose customer data
• Use of customer data to train, improve, or develop artificial-intelligence or machine-learning models
• Whether customer data is anonymized, aggregated, or retained for model training
• Opt-out rights or consent requirements for AI training
• Data-security, confidentiality, or breach obligations
• Indemnification, warranties, liability limitations, or termination rights
• Changes to governing law, dispute resolution, or automatic renewal

AI-Training Analysis

No language addressing AI, machine learning, model training, model improvement, generative AI, aggregated data, or de-identified data appears in the supplied text. Accordingly, there is no identifiable change—either favorable or unfavorable—concerning the use of customer data to train AI models.

Conclusion

The diff appears incomplete and consists only of a page title/navigation fragment rather than the amended legal terms. A meaningful legal comparison requires the full before-and-after text, or a diff containing the marked additions, deletions, and replacements.

Important Changes and Risks

1. Privacy Notice substantially expanded

The replacement adds a comprehensive privacy notice covering:

• Customers, website visitors, end users, and persons acting on behalf of customers.
• Account, payment, device, support, marketing, usage, social-media, analytics, and third-party information.
• International transfers, GDPR/UK GDPR/Swiss rights, California privacy rights, retention, security, and data-subject requests.
• Incorporation of the notice into the Terms of Service and other agreements.

Risk: By accepting the Terms, browsing the website, or using the Services, individuals may be deemed to accept the Privacy Notice. This may create broad consent arguments, although consent may not be valid as the sole legal basis in every jurisdiction.

2. AI-related processing

The most explicit AI change concerns telephone calls:

• Calls may be recorded for customer support, quality assurance, training, and marketing.
• Recordings may be processed with AI tools to generate transcripts, summaries, and notes.
• Call participants will supposedly be informed when recording begins and given an opportunity to opt out.

Risks:

• The notice does not identify the AI providers, processing locations, retention periods, security measures, or whether recordings are used to train third-party or accessiBe AI models.
• “Training” appears as a purpose for recording calls and could be read as employee/service-quality training, but the language does not clearly exclude training machine-learning models.
• There is no express statement that customer data, call recordings, support chats, website content, or end-user data will not be used to train AI models.
• The notice states that all chatbot input is processed, which could include confidential or sensitive information.
• Customers should confirm whether AI vendors may retain prompts, transcripts, or outputs and whether those vendors use data for model training.

3. Broad use of customer and end-user data

The notice permits use of information to:

• Provide, improve, optimize, and support the Services.
• Create aggregated, inferred, anonymized, or pseudonymized data for accessiBe and business partners.
• Analyze website structure, URLs, clicks, interactions, pages viewed, accessibility preferences, and other usage data.
• Share information with service providers, affiliates, business partners, and acquirers.

Risk: “Improve the Services,” “business partners,” and “aggregated” or “anonymized” data are broad terms. The notice does not clearly limit whether customer content or derived data may be used for product development, analytics, benchmarking, or AI training.

4. End-user compliance shifted to customers

For accessWidget end-user data, accessiBe acts as the customer’s processor, while the customer must clearly disclose the processing and obtain required consent.

Risk: Customers bear significant responsibility for notices, consent, and lawful instructions concerning their website users.

5. Other notable changes

• International storage may occur in the United States, EU, UK, Australia, and additional locations at accessiBe’s discretion.
• Marketing communications may be sent for up to six months after certain one-time services.
• Privacy Notice changes become effective when posted; substantial changes receive only ten days’ notice.
• The displayed title/date appears malformed, including “June 28, 2026For,” creating uncertainty about the effective date.

Summary

Available Information

The provided diff contains only the statement:

> “Added approximately 5964 words to the document”

No actual added, deleted, or replaced legal language was included.

Substantive Analysis

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

• What contractual terms changed;
• Whether the customer assumed new obligations or liabilities;
• Whether fees, renewal, termination, warranties, indemnities, or limitation-of-liability provisions changed;
• Whether the provider obtained broader rights to access, use, disclose, retain, or commercialize customer data;
• Whether customer data, prompts, outputs, usage information, or personal information may be used to train, fine-tune, evaluate, or improve AI models;
• Whether the customer can opt out of AI training or revoke consent;
• Whether data is anonymized, aggregated, de-identified, or combined with other customers’ data;
• Whether the provider may use subcontractors or third-party AI vendors;
• Whether the provider offers confidentiality, security, deletion, or data-localization protections for AI-related processing.

AI-Training Risk

The diff does not contain enough information to identify any specific change concerning AI model training. In particular, it is not possible to tell whether the new language:

• Permits training on customer content by default;
• Limits training to de-identified or aggregated data;
• Applies only to service-improvement models or also to general-purpose/commercial models;
• Allows human review of customer data;
• Grants the customer an opt-out or requires affirmative consent;
• Gives the provider ownership or licensing rights in customer inputs or outputs; or
• Continues using data after termination or deletion.

Required Information

Please provide the full redlined text, including the actual additions, deletions, and replacements. If the document is too long, provide at least the sections addressing:

1. Data use and ownership;

2. Privacy and confidentiality;

3. AI, machine learning, or model improvement;

4. Security and retention;

5. Service providers or subprocessors;

6. Intellectual property;

7. Termination and deletion; and

8. Liability and indemnification.

Without the actual language, any legal-risk assessment would be speculative.

The publisher records this document as revised on this date (“Effective as of: June 28, 2026”).

The publisher records this document as revised on this date (“updated on June 4, 2026”).