Following criticism, OpenAI says the company won't use data submitted via its API for “service improvements” like AI model training, unless a customer opts in
As the ChatGPT and Whisper APIs launch this morning, OpenAI is changing the terms of its API developer policy …
TechCrunchKyle Wiggers
Context & Ripple Effects
OpenAI paired the ChatGPT and Whisper API launch with a usage boundary aimed at developers: customer submissions are excluded from service-improvement uses unless the customer affirmatively opts in. That makes data handling part of the API product’s terms, rather than an implicit byproduct of access.
The company later gave consumer ChatGPT users a comparable history opt-out that excludes chats from model improvement, while its advertising plans say conversations will not be exposed to advertisers despite topic-based matching. Across those products, OpenAI is defining distinct permissions for model improvement, product use, and commercialization.
First-order effects
API customers gain a default restriction on OpenAI’s use of their submitted data for model training, with an opt-in path for customers that choose to contribute it.
OpenAI must treat API-submitted data differently from data covered by other product policies, making customer consent a condition of service-improvement use.
Second-order effects
Developers handling sensitive or proprietary inputs have a clearer basis for adopting the new APIs without automatically contributing those inputs to OpenAI’s models.
The policy establishes the permission model later reflected in ChatGPT’s chat-history controls, increasing pressure for OpenAI to keep data-use choices legible across its developer and consumer products.
Third-order effects
OpenAI’s product stack is moving toward purpose-specific data permissions: access to an AI service need not also grant model-improvement rights.
As OpenAI adds data-linked business models while saying it will not expose conversations to advertisers in its topic-matched advertising approach, consent boundaries become a core part of how AI platforms separate customer trust from monetization.
The trend: AI platforms are turning data-use consent into a product-level control, separating model improvement and commercialization from basic service access.
data submitted to the OpenAI API is not used for training, and we have a new 30-day retention policy and are open to less on a case-by-case basis. we've also removed our pre-launch review and made our terms of service and usage policies more developer-friendly.
Don't sleep on the Whisper API announcement - it's incredibly cheap + blazing fast. It's also super accurate, especially for accented or non-native English speakers. Props to @_jongwook_kim and @michpokrass! https://techcrunch.com/...
we frequently don't get it right the first time, but we really do listen and try to rapidly improve. thanks a lot to all of the developers who have given us feedback and patience. more improvements coming! https://twitter.com/...
Cynical take... 1. Release early model 2. Mine queries and responses 3. Improve model until diminishing returns 4. Shut the door behind you https://twitter.com/...
It is vital that any AI platform use great measures to not take data from prompts and build training models. Yet indirectly this may happen, if even unintended. There are ways to build a framework today to assure this. https://twitter.com/...
ChatGPT API now available, 10% the price of our flagship language model & matching/better at any pretty much any task (not just chat). Also released Whisper API & greatly improved our developer policies in response to feedback. We ❤️ developers: https://openai.com/...
we dropped the price of our best text model 10x! amazing work by our research and engineering teams for a long time to make this possible. https://openai.com/...