-
OpenAI
· GPT-4o System Card (PDF)
GPT-4o is programmed with absolute restrictions ('hardcoded OFF' behaviors) that cannot be overridden by any operator or user, including never generating child sexual abuse material (CSAM) or providing serious uplift for weapons of mass destruction....
Why it matters: These absolute prohibitions represent OpenAI's commitment to a floor of safety behavior that no business customer or individual user can disable — but the system card also acknowledges that adversarial prompting can sometimes cause the model to behave inconsistently with its intended restrictions....
View provision →
-
OpenAI
· Privacy Policy (ROW)
OpenAI can use the messages you send to ChatGPT to train and improve its AI systems, unless you turn this off in your account settings....
Why it matters: Sensitive personal information you share in conversations — including health questions, financial details, or private communications — could be used to shape future AI behavior....
View provision →
-
OpenAI
· Privacy Policy (ROW)
OpenAI collects a wide range of data about you automatically, including your messages, files you upload, your location, your device details, and information from other companies....
Why it matters: The breadth of data collection — including conversation content, precise location, and third-party sourced data — creates a detailed profile of users that may be used for purposes beyond what users expect....
View provision →
-
OpenAI
· Privacy Policy (ROW)
OpenAI does not allow children under 13 to use its services and says it will delete any data collected from them if notified....
Why it matters: Without robust age verification mechanisms, children under 13 may access ChatGPT and have their data collected in violation of COPPA, creating legal and safety risks....
View provision →
-
Messenger Kids
· Messenger Kids Terms of Service
A parent or legal guardian must set up and control a child's Messenger Kids account using their own Facebook account, acting as the consent provider for their child's use of the service....
Why it matters: This provision places the legal responsibility for consenting to data collection on the parent, meaning parents are directly accountable for what their child does and shares on the platform....
View provision →
-
Messenger Kids
· Messenger Kids Terms of Service
Meta collects data about how children use Messenger Kids, including messages, activity, and device information, subject to its Privacy Policy which parents should review....
Why it matters: Children's data is among the most sensitive category of personal information, and parents need to understand the scope of what Meta collects in order to make an informed consent decision....
View provision →
-
OpenAI
· Terms of Use (ROW)
If you have a dispute with OpenAI, you must resolve it through private arbitration rather than in court, and you cannot join others in a class action lawsuit against the company....
Why it matters: This clause removes your right to sue OpenAI in court and prevents you from joining with other affected users in a class action, which is often the only practical way to seek compensation for small but widespread harms....
View provision →
-
OpenAI
· Terms of Use (ROW)
OpenAI can use the text, images, and other content you submit to ChatGPT to train and improve its AI systems, unless you turn this off in your account settings....
Why it matters: Your personal conversations, creative work, and other inputs could be used to shape future AI model behavior at a massive scale, raising significant privacy and intellectual property concerns, particularly for sensitive or confidential information you share....
View provision →
-
OpenAI
· Terms of Use (ROW)
You must be at least 13 years old to use ChatGPT, and if you are under 18, a parent or guardian must give permission — though there is no technical age verification mechanism described....
Why it matters: Without a robust age verification system, minors may access the service in violation of these terms, exposing them to AI-generated content risks and creating COPPA liability for OpenAI if children's personal data is collected without verifiable parental consent....
View provision →
-
OpenAI
· Terms of Use (ROW)
If OpenAI gets sued or faces costs because of something you did using their service — including content you created — you must pay OpenAI's legal fees and any resulting losses....
Why it matters: This clause places potentially unlimited financial liability on individual users for their use of ChatGPT, including covering OpenAI's attorneys' fees, which is an unusually broad indemnification obligation for a consumer-facing service....
View provision →
-
OpenAI
· Terms of Use (ROW)
You keep ownership of what you put in, and OpenAI says you own what ChatGPT generates for you — but you're legally responsible if the AI output or your inputs infringe someone else's copyright....
Why it matters: While the ownership grant sounds user-friendly, the warranty that AI outputs don't infringe third-party rights places the copyright infringement risk squarely on users, despite the fact that the AI model — not the user — generates the content....
View provision →
-
OpenAI
· Terms of Use (ROW)
OpenAI limits its financial liability to you to a maximum of $100 or three months of payments, whichever is more, and will not pay for indirect losses like lost income or business disruption caused by AI errors....
Why it matters: If you suffer significant harm — financial loss, reputational damage, or disruption to your business — as a result of ChatGPT errors, hallucinations, or a service outage, OpenAI's legal exposure is capped at a trivial amount that likely won't cover your actual losses....
View provision →
-
OpenAI
· Usage Policies
It is completely forbidden to use any OpenAI product to create sexual content involving children under any circumstances....
Why it matters: This is one of the few absolute, non-negotiable prohibitions in the policy — no operator or user exception applies, and violations will trigger law enforcement referral....
View provision →
-
OpenAI
· Usage Policies
OpenAI's tools cannot be used to help anyone create weapons capable of mass casualties, including biological, chemical, nuclear, or radiological weapons....
Why it matters: This is an absolute prohibition with no operator override, implicating federal criminal law and export controls — violations could expose users and operators to criminal prosecution, not just account termination....
View provision →
-
OpenAI
· Usage Policies
Businesses and developers using the OpenAI API can customize how the AI behaves for their users — making it more or less restrictive — but only within limits OpenAI sets. This means the AI experience varies by platform, and operators bear responsibility for how their users interact with the model....
Why it matters: This clause shifts significant compliance responsibility onto API operators — if a developer's platform enables a prohibited use, the operator bears legal and contractual liability, not just OpenAI....
View provision →
-
OpenAI
· Usage Policies
OpenAI's tools cannot be used to create malware, cyberweapons, or any malicious code intended to cause damage to computer systems or networks....
Why it matters: Given that AI models can generate functional code, this prohibition addresses a specific and serious risk of OpenAI tools being weaponized for cybercrime — violations implicate federal computer fraud law....
View provision →
-
OpenAI
· OpenAI Privacy Policy
OpenAI may use the content of your conversations with ChatGPT and other products to train and improve its AI models unless you opt out through account settings....
Why it matters: This means private, sensitive, or personal information you share with ChatGPT could be reviewed and used to shape future AI behavior, raising significant privacy concerns....
View provision →
-
OpenAI
· OpenAI Privacy Policy
OpenAI shares your personal data with third-party service providers, business partners, and affiliated companies to operate and improve its services....
Why it matters: Your data is not siloed within OpenAI — it is shared with a range of external vendors and partners, expanding the potential exposure of your personal information....
View provision →
-
OpenAI
· OpenAI Privacy Policy
OpenAI's services are not directed at children under a minimum age threshold (typically 13 in the US or 16 in the EU), and parental consent requirements apply for younger users....
Why it matters: If a minor uses OpenAI's products without appropriate consent mechanisms, their data may be collected without the protections required by law, creating risks for both families and OpenAI....
View provision →
-
OpenAI
· OpenAI Terms of Use
OpenAI uses the things you type into ChatGPT and the answers it gives you to train and improve its AI, unless you go into your settings and turn this off....
Why it matters: This is the single most significant data practice in the document — it means every conversation you have with ChatGPT may be used to make the AI better, including any personal or sensitive information you share, and you have to actively opt out rather than opt in....
View provision →
-
OpenAI
· OpenAI Terms of Use
Children under 13 cannot use OpenAI services at all. Teenagers aged 13 to 17 can only use the services if a parent or guardian has given permission....
Why it matters: If a child under 13 uses ChatGPT without authorization, their data may be collected in violation of COPPA, creating legal exposure for both OpenAI and any platform operator that facilitated access; parents should verify their children are not using these services unsupervised....
View provision →
-
OpenAI
· OpenAI Terms of Use
When you use ChatGPT or other OpenAI services, you give OpenAI a broad, worldwide, free license to use, copy, and modify anything you submit — including your prompts and any files you upload — to run and improve the services....
Why it matters: This license grant is broad and covers modification and distribution of your content globally without payment, which means creative work, proprietary business information, or personal data you input into OpenAI services is licensed to OpenAI for operational and improvement purposes....
View provision →
-
OpenAI
· OpenAI Terms of Use
OpenAI can shut down your account at any time and for almost any reason, including if it thinks — but cannot prove — that you have broken the rules....
Why it matters: OpenAI retains essentially unilateral termination rights with minimal procedural constraints, which means paid subscribers can lose access to ChatGPT Plus or API services without a defined notice period, appeal process, or automatic refund entitlement....
View provision →
-
OpenAI
· OpenAI Terms of Use
If OpenAI causes you harm — for example if AI outputs cause financial or reputational damage — your ability to recover money from OpenAI is capped at either what you paid in the last year or just $100, whichever is more....
Why it matters: This liability cap means that even if OpenAI's AI outputs cause significant real-world harm — such as incorrect medical, legal, or financial information acted upon in good faith — your maximum legal recovery is effectively capped at $100 for free tier users, which is far below typical actual damages...
View provision →