On August 7, 2026, the Trump administration released an artificial intelligence testing framework aimed at taking into account the security and cybersecurity risks of new models during evaluation. The details of this framework are currently kept confidential; the White House has no plans to make the full policy available publicly. Only a few technology companies such as OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft were given access to the testing standards and volunteer review processes. This move is seen as a violation of transparency and an advantage for secretive AI companies.
This week, the Trump administration released an artificial intelligence testing framework, incorporating the security and cybersecurity risks of new models into the evaluation. The White House currently keeps the details of this framework confidential, sharing only testing standards with a few technology companies such as OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft. This move is seen as undermining transparency and benefiting secretive AI companies. Although a volunteer review process has been established, the White House does not plan to make the full policy text public.