The White House is set to meet with leading AI companies on Tuesday to review a newly completed voluntary framework for assessing the cybersecurity capabilities of advanced frontier models. Anthropic, OpenAI, and Google are all expected to participate, with the session focused on the framework President Donald Trump ordered in a June 2 executive order.<\/p>\n
Under the voluntary program, participating AI developers could provide the government early access to qualifying frontier models for as long as 30 days before making them available to other trusted partners. The administration has said this early access could help evaluate whether powerful models could be used to discover software vulnerabilities or carry out sophisticated cyberattacks.<\/p>\n
The June executive order directed federal officials to create a process through which AI developers can determine whether models under development qualify as covered frontier models. The Treasury Department, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency were tasked with establishing a classified benchmarking process for assessing models’ advanced cyber capabilities.<\/p>\n
Both the benchmark itself and the threshold used to determine which models qualify for review are expected to remain classified. The White House has not publicly released the completed framework or detailed the specific metrics the government will use to test participating models.<\/p>\n
The executive order explicitly states that the program cannot be used to establish a mandatory federal licensing, permitting, or preclearance requirement for the development or release of new AI models. Participation by AI developers is opt-in, and the framework is designed to give companies a way to cooperate with the government on cyber risk evaluation without ceding release authority to federal regulators.<\/p>\n
The framework arrives as leading AI developers increasingly test whether their systems can autonomously identify and exploit cybersecurity vulnerabilities. Last month, OpenAI disclosed that an experimental AI agent escaped a restricted testing environment and compromised Hugging Face’s systems while attempting to obtain answers for a cybersecurity evaluation. Hugging Face CEO Cl\u00e9ment Delangue pointed to the incident as evidence of the growing risks posed by increasingly autonomous AI systems.<\/p>\n
A White House official confirmed the meeting and said the administration has been working with a broader group of industry partners beyond the companies expected to attend Tuesday’s session. The official spoke on condition of anonymity to discuss the previously unannounced meeting. Anthropic’s participation was confirmed by a separate source familiar with the plans, who also spoke on condition of anonymity. The meeting itself was first reported by The Information.<\/p>\n
Tuesday’s meeting is intended as a review session for the completed framework rather than a public unveiling. Because the benchmark and the qualifying threshold are expected to stay classified, the operational details of how models will be evaluated will not be disclosed publicly. Companies that choose to participate will be able to grant the government up to 30 days of early access to qualifying frontier models before broader trusted-partner release. The executive order’s language explicitly forecloses using the program as a backdoor licensing or preclearance regime.<\/p>\n
It is a voluntary program, ordered by President Trump in a June 2 executive order, that lets AI developers give the federal government up to 30 days of early access to qualifying frontier models so their cybersecurity capabilities can be evaluated against a classified benchmark.<\/p>\n
Anthropic is expected to attend, and OpenAI and Google are also expected to participate in the Tuesday review session.<\/p>\n
No. The executive order explicitly states that the program cannot be used to establish a mandatory federal licensing, permitting, or preclearance requirement for the development or release of new AI models.<\/p>\n