Found in 220 of 352 platforms tracked (62% adoption) · 927 provisions
This clause establishes an affirmative right to notification when automated processes — including profiling — are applied to an individual's Personal Data, providing a check on opaque algorithmic dec…
Users' Personal Data is subject to AI-based processing, which may involve automated decision-making or model inference with distinct risks compared to conventional data handling.
This commitment, scoped to definitions under the Colorado Privacy Act and Oregon Privacy Act, means AWS does not subject consumers to a category of high-stakes automated decision-making those laws sp…
This is a direct commitment that users are not subject to automated decisions or profiling that carry personal consequences, which is significant given the financial nature of Acorns' services.
User interaction data is actively processed by AI systems, meaning automated systems analyze how users engage with Acorns' services for multiple stated and unstated purposes.
Users bear the obligation to ensure their use of AI Features and AI-generated content is independently lawful and compliant with all applicable laws, not merely with ActiveCampaign's platform rules.
The prohibition extends beyond direct use of the Services to cover any derived content, data, or output, broadly restricting AI and ML development activities that touch Adobe's platform.
Users retain a meaningful protection against their Content being used for generative AI training, with one clearly defined exception tied to a voluntary user action.
Users bear full authorization responsibility for transactions executed by AI or automated tools on their behalf, including binding financing terms they may not have directly reviewed.
Affirm's use of customer data to train or utilize AI and machine learning tools means personal information may inform automated systems whose outputs affect users.
The prohibition extends beyond the user to third parties the user might enable, meaning the user bears responsibility for ensuring no downstream recipient of AI Product outputs uses them for model de…
Access by any Agent that fails to identify itself or deviates from section 3 requirements is prohibited, creating a hard compliance condition on all automated interactions with Amazon Services.
The prohibition covers both direct and indirect use and extends to third parties the user controls or permits, closing common circumvention pathways for AI model training.
American Airlines acknowledges a right to challenge purely automated decisions that affect user interests — including profiling across personal, professional, consumer, and credit dimensions — which …
The clause addresses use of potentially biometric data, conditioning it on consent and limiting retention by stating the representations are not stored.
The prohibition covers both automated scraping and any downstream use of Services data for AI development, closing two distinct pathways for large-scale data extraction and model training.
This requirement prevents AI-generated advice or decisions from reaching individuals or consumers without human professional review, establishing a mandatory human-in-the-loop condition.
Anthropic's authority to act exceeds the RSP's mandatory thresholds, meaning development may be paused under a broader set of conditions than those formally defined in the policy.
A revised threshold changes the point at which additional safeguards or restrictions are triggered, directly affecting when Anthropic's RSP obligations activate for automated R&D capabilities.
The requirement creates a formal, documented obligation for Anthropic to affirmatively demonstrate misalignment risk awareness and mitigation at a specific capability level, rather than leaving it to…
The clause establishes a safeguard against purely automated consequential decision-making by requiring that human review remain available, limiting the unchecked use of algorithmic systems against us…
The clause establishes a restriction on a specific and increasingly prominent use of personal data — AI model training — limiting it with respect to private personal data and Apple App Store's founda…
This places an affirmative obligation on users to actively manage AI-related risks, meaning responsibility for harm mitigation is shared with or shifted to the user.
The content of users' photos is analyzed to generate interest inferences that inform targeting, even though these inferences are limited to internal use.
Users' profile information is processed by AI to derive a gender inference that is then applied to targeting, without requiring users to disclose their gender directly.
Users who rely on AI-generated explanations or representations of Best Buy's terms cannot enforce those representations; only the official written terms are legally operative.
The clause is consequential because it defines the extent of Betterment's stated protection: the safeguard rests on third-party assurances rather than Betterment's own direct control over how AI prov…
The clause affirms a present-state commitment against high-stakes automated decision-making and creates a written-notice obligation if BAM's position changes, giving employees a procedural safeguard.
Users retain control over whether their AI interactions are used to train Box's AI models, as this requires affirmative explicit consent.
Automated processing means decisions or flags affecting a user may be generated without human review, and the scope of covered activities is broad.
The provision operationalizes the matching functionality as a core service delivery mechanism rather than an optional feature, and specifies the contractual necessity doctrine as the governing legal …
Users bear responsibility for acting on AI Output that Calendly itself acknowledges may be inaccurate or false, with no apparent recourse against Calendly for harms resulting from such Output.
Customer Data, including potentially sensitive information, leaves Calendly's environment and is processed by third parties whose separate terms and practices govern that data.
Because Cash App disclaims advisory status for its AI features, users cannot treat AI-generated output as professional financial, investment, legal, or tax guidance.
Cohere bars a category of high-stakes automated decision-making, covering domains such as employment, education, and healthcare, where algorithmic outputs could directly affect individuals' life circ…
Automated decision-making may be applied to consequential functions such as account verification without human review, as Coinbase frames this as a regulatory and contractual obligation.
The prohibition covers not only direct use of an AI Agent but also passively allowing or enabling one, expanding the scope of conduct the customer is responsible for preventing.
This clause acknowledges a user right that constrains Copy.ai from making purely automated decisions that legally or significantly affect users without human involvement.
The clause establishes a concrete opt-out from AI-driven grading, giving users a meaningful avenue to invoke the right not to be subject to solely automated consequential decisions.
This clause establishes a recognized right that limits Coursera's ability to make fully automated consequential decisions about users without human involvement.
This commitment means users are not subject to purely automated decision-making with legally significant or similarly weighty consequences from Cursor.
Customer data is shared with an AI services provider, which may involve AI model processing of that data in the United States.
Customer data is shared with an AI services provider, which may involve AI model processing of that data in the United States.
DeepL places the entire EU AI Act high-risk classification and compliance burden on the customer, with no shared responsibility for that assessment or its consequences.
A third-party automated system can result in a subscription request being declined without human review, solely based on Riskified's fraud assessment.
This restriction prevents the Model from being used as the sole decision-maker in contexts with serious legal consequences—such as credit, employment, or contract formation—without any human review.
The definition establishes a dual-consent requirement — consent must run to both the user and Descript — before any voice may be used to train or synthesize an AI Voice.
User information contributes to training automated moderation infrastructure, meaning user-generated data directly informs systems that make enforcement decisions on the platform.
The prohibition explicitly bars a broad range of AI-related activities using Disney+ content, covering the full lifecycle of AI development and deployment.
This prohibition prevents users from leveraging DocuSign's Site—including its data or functionality—as a resource to develop competing or derivative AI systems.
Customer Data—even in de-identified form—may be used to develop AI capabilities, but only where customer consent has been obtained, making the consent mechanism a material safeguard.
Customer Data—which may include sensitive document content—may be used to train AI models under consent terms bundled into the AI-integrated services.
This prohibition broadly covers any software development use of the Services or their content, with AI and machine learning training cited as explicit examples, and requires affirmative written permi…
The designation of AI Ethics as a named subject signals that Dun & Bradstreet's policy framework addresses the ethical dimensions of its AI activities.
The acknowledgment that AI Systems are used in data processing is consequential because it means automated systems may be involved in handling data about individuals.
This prohibition explicitly addresses AI training use cases, preventing extracted game assets or code from being fed into AI models in any training capacity.
This is a direct statement against a high-concern data practice, establishing that automated decision-making with significant consequences—such as profiling for credit, employment, or legal determina…
Asserting a legitimate interest as a legal basis can justify data processing without consent under certain privacy frameworks, broadening Faire's grounds to use data for AI development.
Automated processing of a retailer's historical data directly determines contractual financial conditions—payment terms and invoice deadlines—without human review being specified.
Children's personal data is explicitly excluded from AI training and improvement activities, and this restriction extends to Figma's service providers.
Both Figma and its service providers are prohibited from using children's personal information for any AI development or improvement purpose, providing a categorical protection.
The prohibition covers both the creation and distribution of AI-generated prohibited content and the misrepresentation of synthetic content, establishing two distinct conduct restrictions.
This right gives users a degree of control over how GOAT applies automated processes to their data, though the right is limited to 'certain purposes' that are not defined in this excerpt.
AI assistant interactions — including the user's intent and voluntarily shared content — are within General Motors' potential data collection scope.
The waiver strips commercial AI scrapers of any policy-based or contractual protections they might otherwise invoke to justify their automated access.
AI and machine learning model training using user Inputs and Outputs is the default; users must take affirmative action via account settings to prevent it.
This provision limits fully automated high-stakes decision-making about users, ensuring that decisions with significant legal effects include the possibility of human oversight.
Personal data is being actively used to power AI-driven outputs that directly shape a user's experience, including communications sent on their behalf and candidate matching, making the scope of data…
This right applies specifically to decisions made solely by automated means, giving users a check on algorithmic determinations that define personal, professional, consumption, or credit profiles.
This commitment establishes that Google holds itself to an active, multi-stage standard for reducing harmful AI outcomes rather than simply reacting after the fact.
Automated content analysis is the mechanism by which YouTube Ads produces personalised advertising, meaning ad targeting is driven by machine processing of user content.
This establishes that Google's safety measures are not static but are expressly designed for continuous adaptation, which means the specific safeguards in place may change over time.
This establishes that Google's risk identification is both multi-method and continuous across launch stages, and that findings have a defined downstream effect on policies and frameworks.
This establishes that Google's AI oversight is not limited to pre-launch stages but extends through post-launch monitoring, meaning governance continues after a product reaches users.
This commitment establishes that Google holds itself to a standard of active, multi-stage effort to prevent unfair bias, which is relevant to how AI outputs may affect users.
This commitment establishes that Google's AI systems are intended to be subject to human oversight and feedback processes specifically calibrated to user goals, not solely to internal objectives.
This establishes that Google's responsible AI commitment is not confined to a single stage but applies continuously across every phase of the AI lifecycle.
Users cannot leverage Google's AI-generated outputs to build or train competing or derivative AI systems.
Users who present AI-generated content as human-created are in breach of this prohibition, which has implications for trust and accountability in AI use.
The 'substantially outweigh' standard sets a meaningful threshold above mere net-positive benefit, indicating Google commits to a higher bar before releasing AI products.
Users are placed into ad interest categories not through direct disclosure but through automated inference, meaning category assignment occurs without explicit user input.
This clause establishes a conditional exception to prompt confidentiality: customer prompts are not universally private and may be logged and reviewed when Google Cloud's tools flag potential policy …
Google DeepMind's own acknowledgment that a core safety mechanism has an anticipated expiration point under foreseeable capability growth signals a recognized gap in its current monitoring approach.
The use of 'explore' rather than 'implement' or 'require' indicates this is an investigative or developmental effort rather than an established safeguard, which is consequential for assessing the cur…
This clause establishes that consequential automated decisions affecting individual rights in high-risk domains require human supervision, creating a structural safeguard against unreviewed AI-driven…
The prohibition directly blocks a common use case in AI development, preventing the Customer from leveraging Google Maps Content as training or evaluation data for any AI or machine learning system.
Users cannot rely on Superhuman's AI-generated content as factually accurate and have no warranty claim if such content is wrong.
AI model training is a significant secondary use of collected data; the existence of a user-controlled opt-out mechanism means this use is not unconditional.
Users have a meaningful opt-in or opt-out mechanism over a significant data use—AI model training—rather than it being imposed without recourse.
The requirement that humans are always in control means Talent Matching does not operate as a fully automated decision-making system, preserving human oversight.
Personal Information is explicitly authorized as input for AI model training, meaning user data may shape AI systems beyond the user's individual service experience.
Greenhouse's use of a named third-party auditor for continuous bias audits establishes an ongoing external oversight mechanism.
The prohibition covers both direct and indirect use of Grindr Services for any AI-related development activity, including training and improvement of AI systems.
Groq explicitly disclaims the reliability of its AI outputs and places an affirmative evaluation obligation on the user, meaning users bear responsibility for verifying results before acting on them.
Groq imposes affirmative obligations on users in high-stakes decision-making contexts, meaning failure to conduct risk evaluation or implement safeguards is itself a policy violation, not merely a re…
Groq's explicit acknowledgment that its AI-incorporating Services may carry data protection risks is a formal recognition of potential legal exposure affecting users.
Users are protected from significant solely automated decisions except where Groq can invoke a legal ground and provide prior notification, meaning two conditions must both be met before the protecti…
Groq sets a specific prohibition tied to materiality, harm to individual rights, absence of human supervision, and high-risk domain context—requiring that all four conditions be present for the prohi…
This clause restricts a specific and significant secondary use of personal data by third-party AI providers, offering protection against user data being incorporated into external AI model developmen…
This clause places the entire burden of quality control for AI Outputs on the user, meaning Gusto accepts no responsibility for consequences arising from unreviewed or unapproved AI Outputs.
This clause establishes two distinct opt-out rights that limit Gusto's ability to sell, share, or apply automated decision-making to your personal information.
This clause creates an explicit contractual prohibition preventing Harvey AI from using your data to improve or build AI models.
This disclosure establishes that users cannot rely on Ebb for medical or mental health guidance, which is material for users engaging with Headspace for wellness or mental health purposes.
Users retain an informed opt-in choice before any AI interaction occurs, meaning no AI Feature can be engaged without prior disclosure and user decision.
Recognizing this right means users can challenge how Hilton segments them into consumer sub-groups based on geography, behavior, or demographics for automated purposes.
The clause confirms that benefit determinations can be made entirely by automated systems without human review, conditioned only on user consent.
Material information about profiling and automated decision-making—practices with significant implications for users—is located outside the Privacy Policy, requiring users to consult an additional do…
This clause permits HubSpot to use data customers submit to the platform to improve its AI systems, which has implications for how customer data is processed beyond the immediate service.
Personal data generated through ordinary product use may be used to train AI models, meaning user activity contributes to HubSpot's AI development beyond the immediate service interaction.
OpenAI's role as a sub-processor means data processed through HubSpot AI features may be sent to OpenAI and handled across multiple jurisdictions, including the United States, European Economic Area,…
The prohibition covers a broad range of AI-related activities, barring any use of Hulu's content or services as inputs or resources in AI development pipelines.
A user who generates or uses AI-Generated Content assumes complete legal responsibility for that content, regardless of the AI's role in producing it.
Job seekers' application data is algorithmically evaluated and scored, which may influence hiring decisions without the job seeker initiating or controlling that evaluation.
AI-generated match scores and summary evaluations based on employer criteria may influence hiring decisions, and job seekers are subject to this automated assessment.
Job seekers' application materials may be processed and analyzed by AI tools at the employer's discretion, and job seekers are required to acknowledge this as a condition of use.
This prohibition expressly extends beyond the Services themselves to any data, content, or output derived from them, broadly restricting AI-related uses of anything obtained through the platform.
Automated tools making deactivation decisions mean a user's account can be shut down without human review of the individual circumstances.
Personal data is used not only to deliver current services but to develop and improve Intuit's AI systems, meaning user data contributes to future product capabilities.
The prohibition requires affirmative prior written consent from Kajabi—and potentially from affected users—before any AI or machine learning use of MCP data, creating a dual-consent gate.
User inputs to Khan Academy's AI-Enabled Features are contractually protected from being used by AI providers to train their models.
Khan Academy's AI-Enabled Features are explicitly not designed to handle personally identifiable information, and Khan Academy communicates this limitation directly to users.
This restriction limits one specific high-risk use of Customer Data, providing a defined protection against that data being used to improve external AI systems.
User-generated content and account data may be used to train AI models, meaning user activity directly contributes to the development of Leonardo AI's commercial AI products.
LinkedIn places compliance responsibility squarely on the member for AI-generated content they choose to share, not on LinkedIn as the provider of the AI tool.
Personal data may be incorporated into AI training processes, which can have long-lasting effects on how that data informs model behavior beyond the immediate service context.
Sensitive demographic and financial inferences are derived from profile data and used to target advertising, meaning users may be profiled beyond what they explicitly disclosed.
LinkedIn places sole responsibility for automated ad content and targeting on the advertiser, meaning LinkedIn accepts no liability for outcomes resulting from its own automated recommendations if th…
Customers are granting Luma AI's automated agents authority to act on their behalf, which may have real-world consequences depending on the nature of those actions.
Personal information is processed not only by humans but also by automated systems, which affects the scale and nature of how data is used.
By placing the review obligation on the user, Mailchimp shifts responsibility for the accuracy, legality, and appropriateness of AI-generated content to the user before any use occurs.
Your personal data directly contributes to the development and refinement of AI systems that McDonald's then uses in its service delivery.
Users are protected from legally significant automated decisions being made without separate notice, while profiling for personalization may occur without that safeguard.
The clause establishes a human-oversight requirement for consequential AI-assisted decisions, which bears directly on how Mercury's automated systems may affect users' legal and financial standing.
Users cannot disclaim responsibility for AI Agent conduct by attributing actions to the agent; full responsibility remains with the user who provided the Credentials.
The provision establishes operational requirements for how Microsoft structures AI training practices, including documentation and consent alignment mechanisms. These requirements define the institut…
This provision defines Microsoft's regulatory compliance obligations under EU law and establishes the institutional framework through which the company will implement AI Act requirements across its p…
AI model training on user data means personal information may be incorporated into systems that serve all users, potentially persisting in model weights beyond the original data's lifecycle.
This prohibition directly limits high-stakes automated use of Microsoft's AI services, requiring human oversight before any AI-assisted decision that may consequentially affect key aspects of a perso…
The clause links Microsoft's governing principles directly to the non-discrimination objective across both AI development and deployment.
This establishes a non-discrimination requirement tied directly to Microsoft's named AI principles, making equal treatment and discrimination prevention explicit design obligations.
This provision operationalizes the entity's governance framework for AI system development by establishing fairness and non-discrimination as binding commitments within product design and deployment …
Requiring human oversight means that AI decisions or outputs are not fully autonomous; human review is a stated obligation within Microsoft's responsible AI framework.
Human oversight requirements create governance checkpoints within AI system operations, establishing institutional accountability structures and requiring documented review processes before certain A…
This prohibition places a clear obligation on users to maintain appropriate human oversight whenever AI outputs could consequentially affect any person's legal, financial, life, employment, or human …
The clause establishes a default data usage practice for model training across Mistral's free tier products, with carve-outs for commercial and enterprise offerings. This creates distinct data handli…
The clause establishes a data utilization practice that differentiates service tiers: model training on user interactions occurs only for free-tier users who have not exercised an opt-out mechanism. …
This authorization allows Mistral AI to act as Controller—not just Processor—over Personal Data for AI training, a distinct and broader role that carries different legal consequences, subject to opt-…
The obligation is demand-triggered and open-ended in scope, meaning Mistral AI can at any point require Customer involvement in risk-mitigation activities governed by AI-specific regulation.
The forty-eight-hour window is short and begins at the moment of awareness, placing an immediate and time-sensitive reporting burden on the Customer.
Mixpanel constructs and continuously refines a personal profile about each user, which it uses to predict that user's future behavior toward its Services.
Modal bears no responsibility for the accuracy or fitness of AI Output; all risk of using unreviewed Output falls entirely on Customer.
The prohibition prevents Modal from using Customer Data to improve its own or third-party AI systems without explicit Customer authorisation, protecting the confidentiality and integrity of Customer …
User data may be processed by AI systems, including large language models, subject only to the constraint of applicable law.
The prohibition extends a set of pre-existing restrictions to any machine learning or AI-related activity, preventing users from leveraging Netflix content or service for AI development purposes.
The prohibition extends every prior restriction in the acceptable-use list into AI and machine learning contexts, covering the full lifecycle of model development.
Both human review and automated processing are applied to user-generated content alongside personal information, meaning content is not reviewed solely by algorithms or solely by humans.
Users who leverage Noom's Service for AI or machine learning development are in breach of the terms, which may expose them to account suspension or termination.
User data is used not only to provide the current service but also to build new products and AI/ML systems, which extends the scope of how data is used.
This establishes a specific restriction on how Notion uses a category of user data, preventing its use for AI and ML model development.
This establishes that users are protected from consequential decisions made entirely by automated systems without human involvement, as regulated under certain privacy frameworks.
This clause defines the primary risk categories that OpenAI's governance framework is designed to address, establishing the scope of its safety commitments.
This provision establishes differentiated data practices across user tiers, requiring explicit opt-in consent for certain user categories before their content enters model training workflows, while e…
Explicitly including loss of control establishes that OpenAI's governance framework addresses the risk of AI systems operating outside intended human oversight as a recognized risk category.
This prohibition limits the use of AI-generated Output in high-stakes decisions that directly affect individuals' legal rights or material circumstances, reducing the risk of harm from automated or A…
Content users submit to the Services may be used to improve Oscar Health's AI systems, subject only to the condition that it is de-identified and anonymized before use.
Oscar Health disclaims responsibility for the reliability of AI-generated content and places the full burden of verification on the user, even for content that may not have undergone human review.
Personal information may be used as training or optimization data for Oscar Health's AI systems, even in de-identified or anonymized form, raising questions about the downstream use of derived models.
Users receive wellness guidance from an AI system, meaning the guidance is algorithmically generated from personal health data rather than provided by a human professional.
Personal data may be used to train and improve Oura's AI systems, meaning user data can serve as input for building Oura's proprietary AI capabilities.
Automated systems, rather than humans, are used to make determinations about fraud and risk that may affect the reader's access to PayPal's products and services.
An automated risk determination by PayPal can directly result in denial, termination, or restriction of services, with significant consequences for the user's access to financial services.
Personal Information is not used solely for delivering services to the individual; it may also be used as training data to develop PayPal's AI systems.
Users bear an explicit prohibition against using the tools for consequential decisions, which limits reliance on AI-generated outputs in high-stakes contexts.
Use of the Agentic AI tools automatically exposes a broad category of sensitive Personal Information, including financial data, to those tools.
The clause extends the prohibition beyond direct AI training to indirect use, and covers all content on the platform, creating broad restrictions on how users may engage with Peacock's content.
Users' data is not used to train third-party general AI models, and it is deidentified before being shared, limiting personal exposure in AI contexts.
The strict prohibition on AI training use, combined with a writing requirement for any exception, means users and third parties cannot leverage Peloton's Content for AI development without explicit w…
This prohibition bars use of Perplexity AI's services in high-stakes automated decision-making contexts—such as employment, healthcare, finances, housing, or insurance—where errors could cause materi…
Any harm arising from unverified or inaccurate outputs, or from relying on those outputs as advice, falls on the Customer rather than Perplexity AI.
Customers cannot treat Outputs as authoritative or as professional advice; the burden of accuracy verification rests entirely on the customer.
The absence of a temporal limit on when Pins were posted means that historically posted content remains available for use in training Pinterest's machine learning models.
The customer bears full responsibility for both what is submitted to and what is generated by the AI feature, including legal and contractual compliance.
Because the prohibition covers all Ramp Property and requires prior written consent, Company cannot use any Ramp-provided resource for AI development unless Ramp has explicitly agreed in writing in a…
Because sole responsibility is placed on the user, RapidAPI disclaims accountability for personal data submitted to the chatbot, exposing users to any resulting consequences.
This prohibition limits use cases involving automated determinations about people — such as credit scoring or employment screening — that carry significant real-world consequences.
This clause restricts a high-risk use of AI systems that can determine outcomes—such as credit, employment, or legal status—without human review, which carries significant legal and ethical implicati…
The permission survives termination of the agreement and is not limited to LLM training—'including but not limited to' means other improvement uses are also permitted.
The clause establishes the operational framework for automated eligibility determinations while creating a procedural mechanism for users to challenge those decisions. This addresses regulatory requi…
Consequences for in-game conduct are determined by an automated system, not necessarily by direct human review, with outcomes tied to a comparative behavioral grading process.
Users' in-game communications and behavioral patterns are subject to automated scanning and monitoring, not merely passive logging.
This clause establishes a firm limit on AI autonomy for critical healthcare decisions, ensuring a human professional remains accountable in those determinations.
Users' personal information—including sensitive health data collected by Ro—may be processed by AI systems for multiple business purposes, not solely for direct care delivery.
User data—including potentially sensitive chatbot conversations—may be incorporated into AI model development, which can involve broad internal use and retention beyond the original interaction.
This is a direct commitment that Roblox does not engage in consequential automated decision-making using Personal Information, which addresses a key concern under Brazilian data protection law.
This commitment means users are not subject to consequential automated determinations — such as those affecting legal status or similarly significant outcomes — based on their Personal Information.
This is a direct commitment that Roblox does not engage in consequential automated decision-making using Personal Information, which addresses a key concern under EEA data protection law.
This grant permits Rumble to use submitted Content for AI training and to pass that right to third parties, meaning a creator's Content may be used by outside parties to build or improve AI systems.
This statement addresses a specific GDPR-relevant category of processing; users are told they will not be subject to consequential automated decisions through the Service.
This clause establishes express restrictions on how Salesforce may deploy AI, tying its use to legal permissibility, the Privacy Statement, and Salesforce's own stated standards.
Personal information of U.S. residents can be used to build and improve Samsung's AI systems, which is a broad secondary use of data beyond the original purpose of collection.
This right constrains Segment's use of fully automated decision-making by requiring it to offer human review upon request, and the scope is currently defined by specific account-level decisions.
Personal data is used as training input for AI/ML systems, meaning it informs automated model development beyond direct service delivery.
Automated decisions can directly affect account access, and the clause creates an affirmative obligation on Segment to notify and offer an objection opportunity.
This limits a high-stakes use of personal data—automated decision-making with legal or significant consequences—to situations where the user has consented.
Acknowledging this right establishes that EU and UK users have a formal basis to challenge automated or profiling-based decisions that significantly affect them.
The customer bears all risk associated with relying on AI-generated outputs, with no warranty protection and an explicit exclusion of professional-advice status.
Conversations users have with My AI are used as training data, meaning user inputs contribute to the development of Snapchat's AI systems.
By classifying AI Inputs and Outputs as user-submitted content, Snapchat extends to itself all content licenses users grant elsewhere in the Terms, and holds users responsible for those inputs and ou…
This clause contractually bars users from treating SoFi Coach as a source of professional advice, which limits SoFi's exposure while placing verification responsibility on the user.
This prohibition protects users from consequential automated decisions made without human involvement, a right recognized under major privacy frameworks such as the GDPR.
The AI Terms establish a distinct governance framework for AI Tools, and pre-February 2024 customers or those with applicable Order Forms are bound by it rather than the standard Terms of Service.
Customers who do not fall under the pre-February 2024 or Order Form trigger for the AI Terms are still governed by a defined framework — the main Sourcegraph Terms of Service — for their use of AI To…
This right imposes a meaningful limit on Spotify's use of automated systems by requiring human involvement in decisions that could materially affect users' legal standing or equivalent interests.
Personal and commercial data may be processed by AI systems, meaning sensitive business information can be fed into automated technologies beyond standard data handling.
This prohibition targets AI-enabled psychological manipulation specifically where it impairs informed decision-making and carries likelihood of harm, restricting covert influence operations and dark-…
Social scoring systems can systematically disadvantage individuals based on profiling; prohibiting their use prevents Stability AI technology from being deployed as an instrument of algorithmic discr…
This prohibition directly restricts one of the most invasive applications of AI—using physical or behavioral measurements to infer characteristics that are legally and socially sensitive in most juri…
Automated profiling directly converts personal data into behavioral targeting, meaning a user's own information is used to systematically influence what commercial content they receive.
Automated decision-making and profiling without human intervention can have significant effects on users and triggers specific rights under laws such as the GDPR for EEA residents.
User information is used not just to deliver features but to construct and improve the underlying AI models, including those operated by third-party service providers.
Sensitive categories of data—health and location—may feed into Strava's AI systems, with the scope determined by the user's own privacy settings.
This limitation means users are not subject to consequential automated determinations—such as those affecting legal rights or equivalent interests—by Strava's AI systems.
Personal Data is used as training input for AI systems, which means your data contributes to the development and improvement of Stripe's AI capabilities.
Automated or technological fraud evaluation may be applied to individuals before or during a transaction, affecting how Stripe assesses and responds to those interactions.
The 'at this time' qualifier signals that this human-review requirement is a current practice rather than a permanent policy, leaving open the possibility of future change.
Synthesia explicitly shifts legal compliance obligations for content use and generation to the Customer, insulating Synthesia from liability arising from the Customer's activities.
Customers who are sued for copyright infringement traceable to Synthesia's AI development receive a contractual defence and financial protection from Synthesia.
Users can prevent automated profiling from influencing decisions that carry legal weight or comparably serious consequences for them.
Personal data is used not only to deliver services but to build AI systems, meaning user data contributes to T-Mobile's AI development.
Cloud chat messages are not merely stored but may be algorithmically analyzed for content, meaning message text is subject to automated processing.
This prohibition directly restricts how users may present AI-generated content, requiring disclosure of AI origin and preserving integrity markers such as watermarks.
This right provides a human oversight mechanism for automated decisions that directly affect account access, which is a meaningful procedural protection.
User data is used to train AI/ML systems, meaning personal data may feed ongoing model development beyond immediate service delivery.
Automated decisions can directly affect account access, and the notification and objection requirement is the key procedural protection for affected individuals.
User behavior is continuously analyzed by automated systems against a baseline of typical behavior, meaning algorithmic assessment of behavioral patterns is applied to all users.
Driver and delivery person matching is an automated algorithmic process, not a human-curated one, and is driven by dynamic inputs that include the user's location and destination.
Automated license validation means Uber is programmatically screening credentials at or after onboarding, with account consequences possible if the checks fail.
Matching decisions that affect both riders and drivers are made automatically by algorithms using personal data including real-time location.
Going online to accept work is conditioned on completing a biometric selfie check, affecting drivers' and delivery people's ability to work.
Automated photo verification means Uber is programmatically screening submitted images against authenticity and uniqueness criteria, with potential account consequences for failures.
Driver compensation is determined algorithmically using multiple dynamic variables, meaning pay amounts are not fixed and can fluctuate based on real-time conditions.
Pricing is determined by automated algorithmic calculation rather than a fixed or human-set rate, meaning the price can vary based on multiple dynamic inputs.
Price increases through surge pricing are triggered automatically by supply-demand conditions without any manual intervention or user notification requirement stated.
Automated or AI-driven integration with Unity's Offerings is prohibited by default and requires a specifically authorized access pathway, limiting how developers and automated pipelines can interact …
The notice defines the scope of Unity's data processing disclosures specifically for Generative AI Tools, establishing which AI-related processing activities are disclosed to users.
Users bear direct responsibility for the conduct of all agents, employees, or automated systems they use to access Unity's Offerings, with no insulation for third-party or system-driven violations.
Users and developers cannot use Unity's Offerings or any outputs derived from them as training or validation data for AI or ML models unless they have obtained prior authorization from Unity.
This prohibition directly restricts a growing category of use involving AI development, preventing the Licensed Technology from being fed into generative AI training pipelines.
Automated profiling for consequential decisions is a recognized risk in data protection frameworks, and this commitment addresses that category explicitly, with an extension to known minors.
Paid Pro plan users' content is not used for Model Training unless they affirmatively opt in, and they retain the ability to opt out at any time through specified mechanisms.
Users bear legal responsibility for actions the AI Functionality takes on their behalf, meaning they cannot disclaim those actions even if the AI acted autonomously.
The claim establishes a meaningful limit on how automated analysis of personal information can be applied, excluding its use for consequential individual decisions.
The claim establishes the legal basis Vercel AI relies on for processing user information for AI model development and confirms that user opt-out rights exist alongside this processing.
This clause discloses that customer communications are subject to monitoring, recording, and AI analysis, which affects the privacy of every interaction a customer has with Verizon or its service pro…
Walmart expressly discloses the potential unreliability of AI-generated outputs, which limits user reliance expectations and may affect any claims arising from acting on such outputs.
By excluding AI Feature outputs from the definitions of investment advice, investment research, and trading recommendations, Wealthfront limits the legal and fiduciary weight a user can place on thos…
This clause draws an explicit line between permissible automated document processing and prohibited automated credit decisioning, preserving human involvement in the credit decision.
This clause permits Wealthfront to use Client Personal Information—not just anonymized or synthetic data—across the entire AI development lifecycle, including training and deployment.
The clause provides a direct commitment that automated processing will not be used in ways that produce high-stakes outcomes affecting users, which is a meaningful protection in contexts where such p…
The clause discloses that AI and ML may be applied to user data across a broad set of operational purposes, meaning automated processing of personal data is a permitted practice.
The obligation to provide notice or obtain consent is conditional on legal requirements, meaning no such obligation exists in jurisdictions where law does not mandate it.
User content contributes to the development of Whoop's AI models, though the use is expressly limited to aggregate and deidentified form.
User Content directly contributes to the development and improvement of Windsurf's AI models, meaning content users submit may have a lasting effect on the underlying technology.
This commitment addresses a key concern in data protection law by confirming that no purely automated process produces legally or significantly consequential outcomes for individuals under this Priva…
Workday applies AI-driven inference to a merged dataset to build a profile of your professional interests, which then drives how Workday targets communications to you.
Call and meeting content—not merely metadata—is subject to AI analysis, meaning substantive details of conversations with Workday are processed algorithmically.
This restriction limits how users may deploy AI services in high-stakes contexts and places responsibility on users to ensure human oversight is present.
This clause explicitly extends Yelp's rights over user-submitted content to AI training purposes, a use that goes beyond displaying or promoting content on the platform.
This establishes that user data or interactions processed by Yelp's AI technologies may feed into the training or refinement of AI models, with potential ongoing data use implications.
User-generated content and behavioral interactions are subject to AI-driven analysis, meaning automated systems process personal contributions for personalization and targeting.
Users' browsing histories are used as the basis for targeting, and the targeting is carried out not only by Zillow but potentially by external third-party advertising companies.
This addresses a concern central to AI and algorithmic accountability: whether automated systems can produce consequential decisions about individuals without human review.
Recognizing this right places a constraint on fully automated decision-making that has legal or similarly significant consequences for users.
By placing sole responsibility for hiring decisions on the client, ZipRecruiter limits its own exposure for outcomes resulting from use of its screening functionality.
Users' communications-like Customer Content is expressly excluded from AI model training by Zoom and by third parties through Zoom, providing a defined protection against that specific use.
User personal data is applied not only to eBay's internal AI development but also to the development of third-party AI systems, meaning data flows beyond eBay's own infrastructure.
The clause restricts eBay's use of automated decision-making to three defined bases, providing users with a conditional protection against consequential automated outcomes.
This prohibition prevents the Service from being used as a decision-making tool in contexts where errors could cause serious harm to individuals.
This is a direct commitment by Acorns restricting one specific use of user data, distinguishing it from AI data-processing disclosures that permit broad use.
User content—not just usage metadata—is subjected to machine learning analysis, which has implications for the scope and nature of data processing Adobe performs.
The existence of AI agent deployment at scale signals that automated decision-making processes may operate within user-built applications, which has implications for accountability and oversight.
This commitment directly addresses a significant privacy concern by confirming that Personal Data is not used to subject users to automated decisions or profiling.
This commitment limits Anthropic's use of fully automated decision-making for consequential outcomes, which is directly relevant to rights under privacy frameworks such as GDPR that restrict such pra…
Support responses may be generated by AI rather than a human, which affects the reliability and accountability of the information provided.
Businesses relying on the fraud score to accept or reject transactions may have no recourse to understand or challenge the basis of a given score.
Support ticket contents may include sensitive personal information, and this clause permits that content to be processed by AI systems, including those operated by third parties.
This clause operationalizes the allocation of risk for AI suggestion quality by requiring users to acknowledge specific technical limitations of AI models and establishing that evaluation and risk-be…
Users of Databricks AI features may be bound by terms beyond the base agreement, and those terms are located in external Documentation rather than the agreement itself.
User information may be processed by AI systems, including large language models, which raises distinct considerations about how personal data is handled compared to conventional processing.
The certification signals that Dun & Bradstreet's AI practices have been assessed by a third-party certifying body as of 2024.
Eventbrite can generate new Personal Data about a user beyond what the user directly provides, expanding the profile held about that individual.
It establishes that automated, real-time intervention occurs on user-generated code patterns, meaning the AI system actively acts on user input rather than passively processing it.
Advertisers are subject to both automated and human scrutiny, meaning policy enforcement is not limited to algorithmic detection alone.
Users are guaranteed notice before or during AI-mediated interactions, ensuring they are not unknowingly communicating with an automated system in a context that may affect their healthcare decisions.
Knowing that AWS processes HubSpot AI data in the United States and European Economic Area regions informs readers about where AI-related data processing occurs geographically.
Automated evaluation and prediction of user preferences can influence what job opportunities, recommendations, or content a user is shown, without manual human review.
Message content is processed by automated systems for a range of productivity purposes, conditioned on user settings rather than being off by default unconditionally.
The prohibition covers all Service content including metadata, and extends beyond AI training to any technology designed or intended for identifying natural persons, making the restriction notably br…
Members are subject to ongoing review by both automated systems and human staff, meaning account activity is actively monitored rather than reviewed only on complaint.
The clause limits the weight users can place on Mercury AI output for consequential decisions by explicitly excluding it as a substitute for professional advice.
This requirement imposes an affirmative obligation to deploy specific tooling for AI oversight, not merely to adopt general policies.
This provision creates a formal governance framework that specifies Microsoft's data stewardship obligations in AI development. The operational significance lies in establishing documented standards …
The clause mandates the use of specific tooling for AI oversight, establishing a concrete operational requirement rather than a general aspiration.
The clause signals that model validation is a stated priority in Microsoft's responsible AI approach, linking validation to fairness and real-world accuracy.
Microsoft commits to a structured, multi-layered approach to fairness rather than a single policy, covering principles, practices, and tools.
This commitment establishes a procedural requirement for internal governance of AI system development and deployment, creating organizational obligations to systematically identify and address risks …
This provision establishes Microsoft's operational obligations to provide transparency mechanisms that inform users of AI system involvement and functionality, establishing disclosure requirements as…
This establishes that responsible validation is treated as an important requirement, linking model evaluation to fairness and factual alignment outcomes.
The operational significance lies in the institutional commitment to structured oversight of AI systems across Microsoft's operations. These governance mechanisms establish reporting and accountabili…
The provision operationalizes Microsoft's commitment to disclose AI system functionality and reasoning to users and stakeholders, establishing procedural requirements for documentation and communicat…
Enumerating six specific values establishes the defined framework within which Microsoft's AI product and policy decisions are governed.
The provision establishes transparency regarding data sources used in model training and acknowledges that personal data may be present in training datasets despite filtration practices. This disclos…
The provision establishes the operational basis for Mistral AI's model training methodology and defines the sources from which training data is sourced. It establishes that personal data filtering oc…
Inferred data goes beyond what a user directly provides, meaning Mixpanel's profile of a user may contain conclusions the user never supplied and may not be aware of.
Stating that human values are part of AI policy training implies a normative standard is embedded in the model's governing rules, which affects what behaviors the AI is trained to exhibit.
The claim establishes that live user behavior—including switching and preference signals—is fed back into ongoing model training, meaning user interactions actively shape the routing system.
The claim establishes that users do not directly choose which model processes their query; a router makes that determination based on multiple factors including language in the prompt itself.
Users receive insight into how an AI output was generated at the point it is delivered, rather than having to seek that information separately.
Shein's self-assessment that its automated decision-making is not subject to opt-out rights effectively forecloses that rights avenue for users without independent verification.
This commitment limits the risk of fully automated, high-stakes decisions being made about individuals without human oversight.
User data is not only stored but actively used as training input for algorithmic systems that shape how content is organized on the platform.
The content of support queries—which may contain sensitive personal or business information—is shared with a third-party AI provider for the purpose of improving support quality.
Data involved in AI-powered features of Twilio's product is processed by OpenAI in the USA, a third-party sub-processor with its own data handling practices.
Data processed through Twilio's automated workflows passes through a third-party AI security sub-processor located in the USA, which affects data exposure and jurisdiction.
This clause establishes a limit on how Walmart uses automated profiling, specifically excluding its use for high-stakes decisions affecting consumers' legal rights or equivalent interests.
Wyze applies generative AI inference to personal video footage, meaning content captured by a user's camera is processed by AI models beyond simple storage or playback.
The clause represents Salesforce's stated commitment to balancing innovation with ethical use in the context of AI, which is relevant to how readers may evaluate Salesforce's product development post…
Monitor emails you the same day a platform you choose changes these clauses.
A ai / automated decision-making clause is a provision in a platform's terms of service or privacy policy governing ai / automated decision-making-related rights, obligations, or restrictions.
ConductAtlas tracks 220 platforms with ai / automated decision-making clauses - roughly 62% of platforms in the archive. 548 are classified as high severity.
Severity reflects the magnitude of rights waived, availability of opt-out, breadth of users affected, financial or legal exposure created, and the degree of discretion retained by the platform.