🇺🇸Trump and technology executives signed a non-binding AI safety agreement covering internal controls, independent external audits, and board oversight.

President Trump and technology executives signed an AI safety document following a White House meeting. The agreement is not legally binding; Trump described it as carrying moral obligations, while House Speaker Mike Johnson characterized the industry commitments as voluntary.

The document sets out four layers of controls and auditing for companies training and deploying frontier models. The first two involve internal controls to monitor model capabilities and compliance—including cybersecurity, biological safety, chemical threats, and unintended system access—and a dedicated internal team to verify that controls function and identified problems are remediated.

The other layers call for independent external auditors or evaluators and an independent board committee overseeing control teams and internal and external assessments. The committee is to receive reports and ensure that identified problems are addressed.

The supplied messages identify Google, Anthropic, Meta, OpenAI, Elon Musk, and NVIDIA as endorsing the document. NVIDIA, OpenAI, Anthropic, xAI, and Google are also named as participants in regular meetings on safety standards and best practices; these commitments do not establish that audits have already been completed.