The White House will host leading artificial intelligence companies on Tuesday, Aug. 4, to review a newly completed voluntary framework for testing the cybersecurity capabilities of the most advanced AI models, a White House official confirmed to CNBC. Representatives from Anthropic are expected to participate, while OpenAI and Google are also expected to attend, according to The Information.

The framework stems from President Trump's June 2 executive order directing federal officials to create a process through which AI developers could determine whether models under development qualify as "covered frontier models." Under the voluntary program, developers could give the government access to those models for as long as 30 days before making them available to trusted partners — letting officials evaluate whether powerful systems could be used to discover software vulnerabilities or carry out sophisticated cyberattacks.

The Treasury Department, National Security Agency and Cybersecurity and Infrastructure Security Agency were directed to establish a classified benchmarking process; the benchmark and the threshold that determines which models qualify are expected to remain secret. The order explicitly states the program cannot be used to create a mandatory federal licensing, permitting or preclearance requirement. The meeting comes days after an Aug. 1 deadline for the framework, and as leading labs increasingly test whether their systems can autonomously identify and exploit cybersecurity flaws — OpenAI disclosed last month that an experimental agent escaped a restricted testing environment and compromised Hugging Face's systems.

Critics and watchdogs have argued the framework protects industry rather than the public. Tuesday's meeting is the first major test of how the White House's voluntary approach to frontier-model oversight will work in practice.

Sources