Voluntary commitments to the White House
Seven companies promised to test models for safety, mark generated content and share information about risks.
Why it matters
The first step of American regulation was voluntary and unenforceable, which showed how far the state trailed the pace.
OpenAI, Anthropic, Google, Meta, Microsoft, Amazon and Inflection signed. The commitments covered external testing before release, watermarking generated content and reporting vulnerabilities. No enforcement mechanism was provided. An executive order followed three months later, and binding requirements appeared only in the European regulation.