Every AI Policy Ever · open in the interactive register

SupersededNon-binding · Voluntary code / commitments

Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI

Record US-03

Issued by
The White House (Biden-Harris Administration) with leading AI companies
Jurisdiction
United States
Level · region
National · North America
Published / adopted
2023-07-21
Language
English

Summary

Voluntary commitments by Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI, secured by the White House, to internal and external red-teaming of models before release (including biological, chemical, radiological and cyber risks), information sharing on safety risks with companies and governments, protecting unreleased model weights, third-party discovery and reporting of vulnerabilities, provenance or watermarking for AI-generated audio and visual content, public reports on capabilities and limitations, and research on societal risks. They cover generative models more powerful than the industry frontier at the time.

Safety topics

general-purpose / frontier AIevaluation & testingtransparency & labellingdeepfakes & synthetic mediacatastrophic risk

Scope

Borderline scope Existing scope_note kept: company commitments brokered and published by the White House.

Open question

checker: Only the status is a problem and the run does not change it, so the run's corrections (official_url now the commitments text, fact sheet as secondary, verification note) can be merged while the status question stays with the maintainers (issue #134). Title wording, date and the seven compan Issue #180

The record shows its last checked values until a maintainer decides; the decision is recorded in its history.

Status notes

Announced on 21 July 2023 at the White House. The commitments rest on three principles, safety, security and trust, and the companies intended them to remain in effect until regulations covering substantially the same issues come into force. The White House said it had consulted on them with 20 countries and linked them to the G7 Hiroshima Process and the UK AI Safety Summit. The fact sheet now exists only on the archived Biden White House website. The Joint Commitment on Frontier Responsibilities of 29 September 2026 (US-27) replaced them as the federal government's voluntary framework with frontier AI companies.

Sources

Open the official source https://bidenwhitehouse.archives.gov/wp-content/uploads/2023/07/Ensuring-Safe-Secure-and-Trustworthy-AI.pdf

Secondary source: https://bidenwhitehouse.archives.gov/briefing-room/statements-releases/2023/07/21/fact-sheet-biden-harris-administration-secures-voluntary-commitments-from-leading-artificial-intelligence-companies-to-manage-the-risks-posed-by-ai/

Local copy

Archived. Last attempt 2026-10-18 06:48:43 UTC

Official source (third-party copy, not published) archive/US-03/2026-10-18/official.pdf
Retrieved 2026-10-18 06:48:46 UTC
SHA-256 4f0551712391edcb251e4908276ea66f807ee08ca0e6d8afae8b4125a4bcfe41
Secondary source (third-party copy, not published) archive/US-03/2026-10-18/secondary.html
Retrieved 2026-10-18 06:48:50 UTC
SHA-256 af7c1bc0c4e1d792e7efd2e290bbd7bc7ed67799f4ad89c42215af54478a50df

Verification

B
Another official source, such as an announcement by the issuer, confirms the title, date and status. checked 2026-10-18 · alternative-official-url-fetched
Refresh of 18 Oct 2026 under the rule that level A needs the instrument's own text or an official register. Re-read the two files stored on 11 Oct 2026 on the NARA-archived Biden White House site (both found unchanged by the archiver on 18 Oct 2026): the White House fact sheet of 21 July 2023 and the commitments' own text, a four-page PDF headed 'Ensuring Safe, Secure, and Trustworthy AI' with the list 'Voluntary AI Commitments'. The PDF confirms the eight commitments, their scope, the three principles and the companies' intention to keep them until regulations covering the same issues come into force; it carries no date and does not name the companies. The title wording, the date of 21 July 2023 and the seven companies rest on the fact sheet, an announcement, so they count as alt-official. official_url now points at the commitments' text and the fact sheet becomes secondary_url. 'Superseded' is a judgment resting on record US-27, whose own sources are secondary; no stored file of this record shows the commitments withdrawn or replaced (open question, issue #134). The White House fact sheet of 12 Sep 2023 on eight further companies was requested in an earlier run but could not be stored.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 6 field(s) supported, fixes applied to scope_note.

History

  1. 2026-10-18verification.level: A → B — level computed from the evidence by tools/pipeline.py2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  2. 2026-10-18scope_note: Commitments made by AI companies, which the White House brokered and published; → Existing scope_note kept: company commitments brokered and published by the Whit — pipeline run 2026-10-18-refresh-level-rule-1 (refresh), candidate US-032026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  3. 2026-10-18secondary_url: – → https://bidenwhitehouse.archives.gov/briefing-room/statements-releases/2023/07/2 — the fact sheet, which gives the date and the seven companies, moves here from official_url2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  4. 2026-10-18official_url: https://bidenwhitehouse.archives.gov/briefing-room/statements-releases/2023/07/2 → https://bidenwhitehouse.archives.gov/wp-content/uploads/2023/07/Ensuring-Safe-Se — the record pointed at the White House fact sheet, an announcement; the commitments' own text (the PDF linked from it, stored as document-1 on 11 Oct 2026) is published, so it becomes official_url (extract.md rule 7c)2026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  5. 2026-10-18scope_note: – → Commitments made by AI companies, which the White House brokered and published; — pipeline run 2026-10-17-recheck-north-america-1 (recheck), candidate US-032026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)
  6. 2026-10-18notes: Announced 21 Jul 2023. Never formally withdrawn, but the Biden White House page → Announced on 21 July 2023 at the White House. The commitments rest on three prin — rewritten: the three principles, the intended duration, the 20 consulted countries and the archive banner added from the stored files; the second group of eight companies (September 2023) removed because its source could not be stored; the status explanation shortened2026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)
  7. 2026-10-18summary: Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI committed to i → Voluntary commitments by Amazon, Anthropic, Google, Inflection, Meta, Microsoft — rewritten from the stored PDF: the commitments in the PDF's own terms (biological, chemical, radiological and cyber risks; audio and visual content), and their scope (models above the then-current frontier)2026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)

Related records

White House Frontier AI Accord (2026) United States · Published (non-binding, current)

EO 14110 (Biden AI Executive Order) United States · Revoked / withdrawn

Citation

The White House (Biden-Harris Administration) with leading AI companies. Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI. 2023-07-21. https://bidenwhitehouse.archives.gov/wp-content/uploads/2023/07/Ensuring-Safe-Secure-and-Trustworthy-AI.pdf. Status: Superseded; verified 2026-10-18 (level B). Local copy archive/US-03/2026-10-18/official.pdf retrieved 2026-10-18 06:48:46 UTC, SHA-256 4f0551712391edcb251e4908276ea66f807ee08ca0e6d8afae8b4125a4bcfe41. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record US-03.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.