OpenAI / ChatGPT
7,821 words · ~39 min read · Last updated January 2024
Clause-by-Clause Translation
Original
"By default, we may use Content you provide us to train and improve our models. You may opt out of having your content used to train our models by following the instructions in our Help Center."
Translation
The conversations you have with ChatGPT train future versions of the model by default. The opt-out exists but requires finding it, enabling it, and trusting that the opt-out is applied correctly. The Bureau notes that "by default" is the most important phrase in this clause.
Original
"You may not use the Services to generate content that could be used to create disinformation, election interference, or to impersonate individuals."
Translation
This is in the acceptable use policy, not the terms proper, but the Bureau includes it because of its irony. OpenAI has built one of the most capable content generation tools in history and prohibits its use for content generation at scale. The prohibition is enforced by automated classifiers. The Bureau notes that automated classifiers have failure modes.
Original
"We do not guarantee that the Services are accurate, complete, reliable, current, or error-free. OpenAI is not responsible for the accuracy or completeness of any information provided by the Services."
Translation
ChatGPT can be confidently wrong. OpenAI's terms note this explicitly. If you use ChatGPT output in a professional context, make a decision based on it, or rely on it for health or legal information, and that output is wrong — OpenAI has disclaimed responsibility for this outcome.
Original
"To the maximum extent permitted by law, OpenAI's total liability to you for any claims arising out of these terms is limited to the greater of $100 or the amount you paid OpenAI in the past 12 months."
Translation
If ChatGPT causes you harm — through wrong information, data exposure, or any other mechanism — and you have a free account, OpenAI's maximum legal liability to you is $100. One hundred dollars. The Bureau has noted this figure and would like you to sit with it.
Original
"OpenAI may update these Terms from time to time. We will notify you of significant changes. Your continued use of the Services after the effective date of any changes constitutes your acceptance of the new Terms."
Translation
Standard clause. The Bureau includes it because it is standard. Your continued use of a product is legally equivalent to signing a contract. Every morning you open ChatGPT, you are re-signing.
What You Actually Agreed To
That your conversations train OpenAI's models by default (opt-out available)
That OpenAI is not responsible for the accuracy of any information the model provides
That OpenAI's liability to you is capped at $100 (for free account users)
That continued use constitutes ongoing acceptance of any updated terms
That disputes are resolved through individual arbitration, not class actions
That OpenAI can modify or discontinue the service at any time
Bureau Summary
"OpenAI's terms are, relative to the field, fairly direct. This is a company whose entire product is trained on data, and their terms reflect that. The Bureau notes that the questions you asked ChatGPT — the personal ones, the professional ones, the 3am ones — are covered by these terms. The Bureau also notes that OpenAI offers an opt-out mechanism, which is better than most."
— Bureau Translation Service