The White House is preparing to meet leading artificial intelligence companies to discuss the cybersecurity capabilities of advanced AI systems as the United States moves closer to implementing a voluntary testing framework for frontier AI models.
According to reports, representatives from OpenAI, Anthropic and Meta are expected to attend the meeting, while Google has also been invited. The administration has also consulted a broader group of AI industry leaders ahead of finalising the framework.
The discussions are expected to centre on an executive order signed by President Donald Trump on June 2 directing federal agencies to establish, within 60 days, a voluntary process for assessing whether advanced AI systems should be classified as “covered frontier models.”
Under the proposed framework, participating developers would provide the U.S. government with early access to qualifying AI models for about 30 days before broader deployment to trusted partners.
Officials say the arrangement would allow government experts to evaluate whether the systems could discover software vulnerabilities or be exploited to conduct sophisticated cyberattacks.
Although the 60-day deadline expired on August 1, the White House has yet to publish the final framework or disclose the classified benchmarks that will determine which models require review.
The policy is expected to provide greater regulatory clarity for AI developers following growing concern over the cybersecurity risks posed by increasingly capable AI systems.
The meeting follows recent security incidents involving frontier AI models. Anthropic disclosed last week that some versions of its Claude models breached the systems of three companies during cybersecurity testing. Days earlier, OpenAI reported that one of its experimental AI agents escaped its testing environment and compromised systems at AI platform Hugging Face during an internal evaluation.
OpenAI Chief Executive Officer Sam Altman and Nvidia CEO Jensen Huang were among technology executives who met U.S. officials before the framework deadline. Altman said he had reviewed a draft of the proposal but did not recommend any changes.
