Key facts
- Two dozen tech firms voluntarily committed to implementing White House-recommended controls for AI safety.
- The accord includes commitments to independent safety audits testing cybersecurity, biosecurity, chemical threats, and unintended AI actions.
- Signatories include leaders from Anthropic, OpenAI, xAI, Nvidia, Meta, and Alphabet/Google.
- The agreement is described as "morally binding" by President Trump, though it carries no legal weight.
- Some firms had already committed to external audits and government safety testing prior to this accord.
- The accord suggests recommendations may eventually be codified into law.
President Donald Trump and leaders from two dozen major technology firms have agreed to a voluntary accord for AI developers to self-police frontier AI risks. The agreement, described as "morally binding" by Trump, involves companies committing to implementing recommended controls, including independent safety audits, and regularly meeting to discuss best practices and set common AI safety standards.
Key focuses for the external reviews will include risks related to cybersecurity, biosecurity, chemical threats, and unintended actions by AI models. Signatories include leaders from Anthropic, OpenAI, xAI, Nvidia, Meta, and Alphabet/Google. Some of these firms had already committed to external audits and government safety testing prior to this accord.
While the agreement carries no legal weight, Trump stressed that trusting AI firms to regulate themselves is the best path forward for the US. The accord suggests that recommendations may eventually be codified into law. Critics, however, have expressed concerns that self-policing may not adequately address AI safety issues, with some accusing Trump of favoring industry interests over stricter regulatory guardrails.