-
ElevenLabs
· ElevenLabs Usage Policy
You cannot use ElevenLabs to create audio that harasses, threatens, or abuses people, facilitates illegal activity, involves child sexual abuse material, or otherwise breaks the law....
Why it matters: This provision establishes the outer limits of permitted use and places responsibility on users for ensuring their generated content is lawful; the prohibition on child sexual abuse material in particular has mandatory reporting and compliance dimensions that go beyond standard platform terms....
-
Runway
· Runway Usage Policy
Runway prohibits any content involving child sexual abuse, automatically reports it to NCMEC and global law enforcement, and permanently bans all accounts associated with such content....
Why it matters: This provision establishes Runway's most serious enforcement mechanism, an indefinite account suspension with no described appeal path, and triggers mandatory law enforcement referral for CSAM violations....
-
Runway
· Runway Usage Policy
Runway prohibits any attempt to use its tools to create or alter intimate images of real people without their consent, commonly referred to as deepfake pornography or NCII....
Why it matters: Non-consensual intimate imagery is an expanding area of criminal and civil law in the U.S. and internationally, and this provision prohibits the use of Runway's platform for such purposes explicitly....
-
Runway
· Runway Usage Policy
Within Runway's Characters and Game Worlds products, users cannot create AI characters based on minors' faces or voices, cannot target content at children, and cannot create characters that simulate professional advice or therapy....
Why it matters: These restrictions reflect both child safety obligations and emerging concerns about AI systems being deployed as professional advisors without appropriate oversight; they have significant implications for enterprise users building AI characters or interactive experiences....
-
Cursor
· Cursor Data Use & Privacy Overview
If Privacy Mode is off, Cursor states it may use your codebase data, prompts, editor actions, and code snippets to train its AI models, and may share prompts and limited telemetry with third-party model providers you select....
Why it matters: This provision establishes the full scope of data use when Privacy Mode is disabled, authorizing collection and use of codebase data, prompts, and editor actions for AI model training and disclosure to third-party model providers....
-
Monitoring
These provisions have changed before.
Monitor includes same-day alerts, structured change summaries, and monitoring for up to 25 platforms.
-
OpenAI
· OpenAI Data Processing Addendum
OpenAI's standard API services are not set up for healthcare data protected under US law. If you want to use the API with patient health data, you must sign a separate legal agreement with OpenAI first....
Why it matters: This provision places the compliance burden on the operator to identify when HIPAA applies to their use case and to execute a BAA before submitting any protected health information. Using the API with PHI without a BAA in place would constitute a potential HIPAA violation by the operator....
-
Stability AI
· Stability AI Acceptable Use Policy
The policy prohibits using Stability AI's models to generate any sexual content involving minors, regardless of whether the output is photorealistic or stylized....
Why it matters: Generation of CSAM is a criminal offense in virtually all jurisdictions, and this prohibition establishes that any such use constitutes an immediate and absolute violation of the policy with no exceptions or contextual carve-outs....
-
Stability AI
· Stability AI Acceptable Use Policy
The policy prohibits using Stability AI's models to create synthetic media, including realistic images, video, or audio of real people, that is designed to deceive viewers about its artificial origin or to misrepresent a real person's statements or actions....
Why it matters: This prohibition covers a category of AI-generated content that is increasingly the subject of specific legislation in multiple jurisdictions, and it establishes that creating non-consensual intimate imagery or politically deceptive deepfakes using Stability AI's tools violates the policy....
-
Stability AI
· Stability AI Acceptable Use Policy
The policy prohibits using Stability AI's models to assist in developing weapons of mass destruction, cyberweapons, or to conduct attacks on critical infrastructure such as power grids, water systems, or financial systems....
Why it matters: This prohibition addresses national security-adjacent use cases and establishes that any attempt to use the AI for weapons development or infrastructure disruption is a policy violation subject to immediate termination and potential referral to law enforcement....
-
Stability AI
· Stability AI Acceptable Use Policy
The policy prohibits using Stability AI's models to generate content that harasses, threatens, or targets individuals, including generating intimate or sexual imagery of real people without their consent....
Why it matters: Non-consensual intimate image generation is criminalized or subject to civil liability in a growing number of jurisdictions, and this prohibition establishes that producing such content using Stability AI's tools constitutes a policy violation regardless of whether the output is photorealistic....
-
Hugging Face
· Hugging Face Content Policy
The policy prohibits a specific list of harmful content types including any sexual content involving minors, non-consensual sexual content, content that reveals private information about others without permission, and content that infringes intellectual property rights....
Why it matters: These enumerated categories establish the baseline content standards that all users must comply with when uploading ML models, datasets, community posts, or other content to the platform; violations are subject to immediate enforcement action....
-
Jasper AI
· Jasper Privacy Policy
Jasper may use what you type into the platform, including your questions to the AI and the responses it generates, to improve its AI systems, unless you opt out....
Why it matters: Business users submitting confidential, proprietary, or client-related content as prompts should be aware that this material may be used to train Jasper's AI models unless they actively exercise the opt-out, which has practical implications for data confidentiality obligations....
-
OpenAI
· OpenAI Usage Policies
OpenAI absolutely prohibits using any of its tools or models to generate sexual content involving minors, with no exceptions regardless of context, operator permissions, or stated purpose....
Why it matters: This is a hard, unconditional restriction that applies to every user and operator without exception; violation would constitute both a policy breach and potentially criminal conduct under laws in most jurisdictions....
-
OpenAI
· OpenAI Usage Policies
OpenAI prohibits using its tools to meaningfully assist anyone attempting to develop biological, chemical, nuclear, or radiological weapons capable of causing mass casualties, regardless of the stated purpose....
Why it matters: This provision addresses one of the highest-risk potential misuses of generative AI, and its scope covers not just direct weapons synthesis but any assistance that provides 'serious uplift' — a term that implies a meaningful capability increase — to someone pursuing such weapons....
-
OpenAI
· OpenAI Usage Policies
OpenAI prohibits using its models to generate cyberweapons, malware, or other malicious code that could cause significant damage, distinguishing this from permissible cybersecurity research and defensive security work....
Why it matters: This provision applies to all users and operators and covers generation of offensive cyber tools, though the document implicitly acknowledges a distinction between prohibited offensive tool creation and permitted defensive security research — a distinction that may not always be clear in practice....
-
OpenAI
· OpenAI Usage Policies
OpenAI prohibits using its tools to produce content that encourages or celebrates real-world violence, terrorism, or attacks on infrastructure, and prohibits providing operational planning assistance to those conducting such attacks....
Why it matters: This prohibition applies to both direct incitement and operational assistance, covering a range of content from propaganda to logistics support for violent acts, and applies regardless of whether the requester claims an educational, journalistic, or creative purpose....
-
Midjourney
· Midjourney Community Guidelines
Creating, uploading, sharing, or attempting to distribute any content that sexualizes minors, whether real or AI-generated, is absolutely prohibited on Midjourney....
Why it matters: This provision reflects a legal obligation under US federal law (PROTECT Act, FOSTA-SESTA adjacent obligations) and similar laws globally, and violations carry criminal liability independent of platform enforcement....
-
Mistral AI
· Mistral AI Commercial Terms
Mistral AI generally does not use your business data to train its AI, but there are five exceptions, including when you use Labs Models, submit Feedback, or use free or certain paid subscriptions without opting out. For those exceptions, you grant a permanent, irrevocable license to use your data for training....
Why it matters: The license granted for training purposes under the listed exceptions is perpetual and irrevocable, meaning that once data falls within one of these categories, Mistral AI's right to use it for training cannot be withdrawn even if the Customer later terminates the agreement....
-
Mistral AI
· Mistral AI Commercial Terms
When you use experimental Labs Models, your data and outputs are used for AI training by default, and any opt-out settings you configured for other Mistral AI products do not carry over. You must separately activate zero data retention within Labs Models to prevent this....
Why it matters: This provision creates a product-siloed consent architecture where a business that has opted out of training for its primary Mistral AI products may unknowingly contribute training data through Labs Models unless it takes a separate, affirmative action per product....
-
Mistral AI
· Mistral AI Commercial Terms
If your business submits any feedback about Mistral AI's products, Mistral AI can use that feedback, plus any data and outputs associated with it, in any way it chooses, permanently and without any payment or IP restriction....
Why it matters: The Feedback license extends beyond the feedback itself to include the Customer Data and Outputs associated with that feedback, and explicitly excludes these from confidentiality protections, meaning that sensitive business data submitted alongside feedback may be used without restriction....
-
Mistral AI
· Mistral AI Commercial Terms
Your organization bears full legal responsibility for everything that happens through your account, including what your end users do and whether you have obtained the necessary consents from them before giving them access to Mistral AI products....
Why it matters: This provision places the full burden of End User compliance, including consent collection, on the commercial Customer rather than Mistral AI, meaning that if an End User violates the terms or applicable law, the Customer is the liable party....
-
Character.AI
· Character.ai Community Guidelines
Character.AI states it can permanently close your account or refer you to law enforcement if it determines your behavior is a serious violation of its rules....
Why it matters: The provision asserts broad enforcement discretion including law enforcement referral without defining what constitutes a 'serious case' or describing any notice, appeal, or procedural safeguard for affected users....
-
Character.AI
· Character.ai Community Guidelines
Character.AI prohibits all sexually explicit content, child sexual abuse material, grooming behavior, sexual extortion, pornography, and nudity on the platform....
Why it matters: The explicit prohibition on child exploitation material and grooming reflects mandatory legal obligations under federal law and directly implicates the platform's CSAM reporting duties to the National Center for Missing and Exploited Children....
-
Character.AI
· Character.ai Community Guidelines
Character.AI prohibits content that promotes, glorifies, or encourages self-harm, suicide, or eating disorders, including extreme fitness content and body shaming....
Why it matters: This provision has direct relevance to user safety, particularly for minor users, and engages ongoing public and regulatory scrutiny of social and AI platforms regarding mental health content moderation obligations....
-
Cohere
· Cohere Enterprise Data Commitments
Cohere states that it will not use the data you send through its platform (such as prompts and documents) to train its AI models unless you specifically agree to allow it....
Why it matters: This provision directly addresses a common concern for enterprise customers deploying AI: whether proprietary business data submitted as prompts or documents could be incorporated into shared model training. The document states this requires explicit opt-in rather than an opt-out....
-
Cohere
· Cohere Enterprise Data Commitments
Cohere states that enterprise customers can choose where their data is stored and processed geographically, which is important for organizations with data localization requirements....
Why it matters: Data residency options are operationally significant for enterprises subject to data localization laws or contractual requirements restricting cross-border data transfers. The availability of these options depends on Cohere's infrastructure and the specific regions offered....
-
Skillshare
· Skillshare Terms of Service
If you have a dispute with Skillshare, you agree to resolve it through private arbitration rather than in a public court, and you give up the right to join a class action lawsuit or have a jury decide your case. This applies even to disputes you may have had with Skillshare before you agreed to these Terms....
Why it matters: The retroactive application of this arbitration clause, covering claims that arose before the effective date of these Terms, is operationally distinct from many platforms that limit arbitration to future disputes, and significantly narrows the legal avenues available to users with existing grievances....
-
Skillshare
· Skillshare Terms of Service
You give up your right to band together with other Skillshare users to bring a joint lawsuit or arbitration against the company, and you also waive your right to a jury trial for any dispute....
Why it matters: The class action waiver means that even if many users are harmed by the same Skillshare practice, each person must pursue their claim individually, which can make it economically impractical to seek redress for smaller harms....
-
Mistral AI
· Mistral AI Additional Product Terms
When you connect external services like MCP Servers to Mistral AI products, you authorize Mistral AI to send your data to those services, and Mistral AI takes no responsibility for any data loss or disruptions that result from using them....
Why it matters: The agreement authorizes Customer Data to be transmitted to third-party services upon customer configuration, while fully disclaiming Mistral AI's liability for any resulting data loss, which shifts significant operational and legal risk to the customer....
-
Mistral AI
· Mistral AI Additional Product Terms
Any organization that regularly publishes news or editorial content is restricted from using Mistral AI's verified news features without prior authorization from Mistral AI, and unauthorized use can result in immediate account suspension without a refund....
Why it matters: The agreement defines Media Customers broadly to include any entity that disseminates information for editorial purposes, and imposes account suspension as an automatic consequence of unauthorized news feature use, creating significant operational risk for media organizations....
-
Mistral AI
· Mistral AI Additional Product Terms
Customers using Mistral AI's audio and voice cloning features are prohibited from cloning voices without explicit consent or using the tools to impersonate, deceive, or generate harmful content, and must disclose AI-generated audio where the law requires it....
Why it matters: The terms establish explicit prohibitions on non-consensual voice cloning and deceptive audio generation, and require legal disclosure of AI-generated audio content, while fully disclaiming Mistral AI's liability for any non-compliant use by the customer....
-
DeepSeek
· DeepSeek Open Source License
Businesses and developers can use the DeepSeek-V3 model commercially for free, but only if their platform or product has fewer than 100 million monthly active users. Organizations that exceed that threshold must obtain a separate written license directly from DeepSeek....
Why it matters: This clause creates a hard commercial licensing gate tied to platform scale. Any organization that crosses the 100 million MAU threshold without a separate agreement is operating outside the terms of this license, and the clause grants DeepSeek sole discretion over whether to issue that license....
-
DeepSeek
· DeepSeek Open Source License
You cannot use the DeepSeek-V3 model or its outputs to build other AI models that you distribute or license to others. You may only use it to build models strictly for internal business use that are not shared externally....
Why it matters: This clause prohibits using DeepSeek-V3 or any of its generated outputs as a basis for training or developing external AI models. This restriction goes beyond standard open-source licensing and could affect AI research, fine-tuning pipelines, and commercial AI product development workflows that rely on this model as a foundational component....
-
DeepSeek
· DeepSeek Open Source License
Users cannot use DeepSeek-V3 to help build weapons of mass destruction, generate child sexual abuse material, engage in illegal activities, or violate others' rights. An external Acceptable Use Policy is also incorporated into the license by reference....
Why it matters: The prohibited use clause establishes absolute restrictions on harmful applications of the model. The incorporation of an external Acceptable Use Policy by reference means the full scope of restrictions extends beyond the license text itself, and users are bound by that policy without it being reproduced here....
-
OpenAI
· OpenAI Business Terms
If you have a legal dispute with OpenAI, you generally must resolve it through private arbitration rather than in a public court, and you cannot join a class action lawsuit against OpenAI....
Why it matters: This provision requires users to waive their right to a jury trial and class action participation, meaning individual disputes with OpenAI must proceed through a private arbitration process governed by AAA Consumer Arbitration Rules....
-
OpenAI
· OpenAI Business Terms
You cannot sue OpenAI as part of a group or class action lawsuit; any legal claim must be brought individually, whether in arbitration or (if the waiver is unenforceable) in court....
Why it matters: The class action waiver means that even if many users experience the same harm, each must pursue claims individually, which may make low-value individual claims economically impractical to pursue....
-
OpenAI
· OpenAI Business Terms
When you use OpenAI's services, you give OpenAI permission to use what you type in (prompts) and what the AI generates in response (outputs) to operate and improve its services....
Why it matters: This license authorizes OpenAI to use the content of user conversations, including potentially sensitive or proprietary information submitted as inputs, for purposes including service improvement, which may have implications for users submitting confidential business or personal information....
-
OpenAI
· OpenAI Business Terms
If something goes wrong and OpenAI is at fault, the most you can recover from OpenAI is the greater of what you paid them in the past year or $100, and you cannot claim compensation for indirect harms like lost profits or consequential losses....
Why it matters: This cap means that even if a user suffers significant harm attributable to OpenAI's services, the agreement limits the maximum financial recovery to $100 for free-tier users or the amount paid in the prior 12 months for paying subscribers, and excludes indirect or consequential damages entirely....
-
Cohere
· Cohere SaaS Agreement
Unless you opt out, Cohere may use the text you send to and receive from the API to train and improve its AI models....
Why it matters: The agreement authorizes use of customer-submitted inputs and model outputs for model training by default; enterprise customers transmitting confidential, regulated, or sensitive data should confirm their opt-out status before using the API in production....
-
Cohere
· Cohere SaaS Agreement
If you use Cohere's API to process personal data, a separate Data Processing Addendum applies that sets out additional privacy and data protection obligations....
Why it matters: The DPA structure is the primary mechanism through which GDPR, CCPA, and other data protection obligations are operationalized in the agreement; enterprise customers processing personal data through the API must ensure the DPA is executed and that its terms are consistent with their privacy compliance obligations....
-
Meta
· Llama API Terms of Service
Developers are prohibited from selling, licensing, or transferring user data obtained through Meta's platform to advertising networks, data brokers, or monetization services. They can only collect data that is strictly necessary for their app to work....
Why it matters: This provision directly restricts how developers can monetize user data obtained through Meta's APIs, which affects consumers whose data is collected by third-party apps using Facebook Login or similar integrations....
-
Meta
· Llama API Terms of Service
Developers are responsible for writing and publishing their own privacy policy and for getting users' consent before collecting their data. The privacy policy must be at least as protective as Meta's own policies....
Why it matters: This provision places independent legal and operational obligations on developers to maintain compliant privacy disclosures and consent mechanisms, meaning that failures in these areas are the developer's responsibility rather than Meta's....
-
Perplexity AI
· Perplexity Acceptable Use Policy
Generating, sharing, or enabling content that sexually exploits children is strictly prohibited and likely illegal under applicable law....
Why it matters: This is an absolute prohibition aligned with US and international law; violations may expose users to criminal liability beyond account termination....
-
Perplexity AI
· Perplexity Acceptable Use Policy
You cannot use Perplexity to gather other people's personal information without permission, build stalking or tracking tools, or conduct surveillance without consent....
Why it matters: This provision addresses AI-enabled privacy violations, including the use of generative AI to build surveillance or data harvesting tools targeting individuals without their knowledge....
-
Perplexity AI
· Perplexity Acceptable Use Policy
You cannot use Perplexity to generate content that promotes violence, terrorism, or hate-motivated harm against individuals or groups....
Why it matters: This prohibition addresses the risk of generative AI being used to radicalize users or provide operational support for violent acts, and aligns with both platform policy norms and legal requirements in multiple jurisdictions....
-
OpenAI
· OpenAI Service Terms
Instead of suing OpenAI in court, this clause requires you to resolve most disputes through private arbitration. You can opt out within 30 days of agreeing to the terms by sending written notice to OpenAI....
Why it matters: Binding arbitration means disputes are resolved outside the court system, typically by a private arbitrator, which affects the procedural rights and remedies available to users compared to court proceedings....
-
OpenAI
· OpenAI Service Terms
This clause means you cannot join or lead a class action lawsuit against OpenAI; any legal claim must be pursued individually....
Why it matters: The class action waiver limits users' ability to pool claims with other affected individuals, which is particularly relevant for disputes involving small individual harms that may only be practical to pursue collectively....
-
OpenAI
· OpenAI Service Terms
When you use ChatGPT or other OpenAI services, the company can use your conversations to improve its AI models unless you turn off this option in your account settings....
Why it matters: This provision authorizes OpenAI to use the content of your interactions, including prompts and responses, as training data for AI model development, subject to an opt-out that users must actively exercise....
-
OpenAI
· OpenAI Service Terms
OpenAI's maximum financial responsibility to you for most claims is limited to either what you paid them in the past year or $100, whichever is more....
Why it matters: This provision sets a low ceiling on what OpenAI would owe in most legal disputes, including those involving service failures, data handling issues, or other harm; for free-tier users, the cap is $100 regardless of the nature of the harm....
-
OpenAI
· OpenAI EU Terms of Use
When you submit content to OpenAI services, the agreement grants OpenAI a license to use that content to provide and potentially improve its services, which may include AI model training depending on account settings....
Why it matters: The scope of this content license determines whether personal data or proprietary information submitted as inputs can be used for purposes beyond delivering the immediate service response, which engages GDPR lawful basis and purpose limitation requirements....