Prompt API

[domenic]

こんにちは TAG-さん!

I'm requesting an early TAG design review of the Prompt API.

This is a JavaScript API for directly prompting an AI language model, including text, image, and audio inputs.

Although the Chrome built-in AI team is exploring purpose-built APIs for some use cases (#948, #991), we are also exploring a general-purpose "prompt API" which allows web developers to prompt a language model directly. This gives web developers access to many more capabilities, at the cost of requiring them to do their own prompt engineering.

• Explainer: https://github.com/webmachinelearning/prompt-api/blob/main/README.md
• User research: none (although we have developer-facing research)
• Security and Privacy self-review: https://github.com/webmachinelearning/prompt-api/blob/main/security-privacy-questionnaire.md
• GitHub repo: https://github.com/webmachinelearning/prompt-api
• Primary contacts:
  • Domenic Denicola (@domenic), Google, spec editor
  • Etienne Noël (@etiennenoel), Google, team lead
• Organization/project driving the design: Google
• This work is being funded by: Google, Microsoft
• Incubation and standards groups that have discussed the design:
  • This API has been discussed at the majority of Web Machine Learning CG meetings in 2025, the minutes for which can be found at https://github.com/webmachinelearning/meetings/tree/main/telcons
• Standards group(s) that you expect to discuss and/or adopt this work when it's
  ready: W3C Machine Learning Working Group
• Multi-stakeholder feedback:
  • Chromium comments: prototyping
  • Mozilla comments: Prompt API mozilla/standards-positions#1213 and some comments in the Web ML CG minutes listed above
  • WebKit comments: ML Prompt API WebKit/standards-positions#495
  • Developer comments: some of them are collated in https://github.com/webmachinelearning/prompt-api/blob/main/README.md#stakeholder-feedback
• Major unresolved issues with or opposition to this design: There are a lot of enhancement requests on the issue tracker. There are a couple of interop concern issues there as well. It all seems manageable at this point, though.
• I have reviewed the TAG's Web Platform Design Principles

You should also know that...

Although this feature exposes an on-device language model, the language model is never trained on, and does not have access to, any local user-specific data. (Sometimes people seem to be confused by this, so we're trying to include this disclaimer in as many places as possible.)