White House to Review New Model Testing Framework with AI Companies on Tuesday

The White House will discuss a voluntary testing framework with major AI companies on Tuesday to evaluate the cybersecurity capabilities of advanced AI models. Anthropic, OpenAI, Google, and others are reportedly expected to attend, and the administration has been working with a broader range of industry partners.

Under this program, established following President Donald Trump's June 2 executive order, participating companies may provide the government with up to 30 days of early access before releasing models publicly or to trusted partners. The government and companies will assess whether models can discover software vulnerabilities or carry out sophisticated cyberattacks.

The Treasury Department, the National Security Agency (NSA), and the Cybersecurity and Infrastructure Security Agency (CISA) will establish classified benchmarks to evaluate models' advanced cyber capabilities. The benchmarks and application criteria will remain classified, and the administration has stated that this program cannot be used as a mandatory licensing or pre-approval system for the development or release of new AI models.