The United States government is pressuring Meta to voluntarily submit its advanced artificial intelligence models for federal security reviews. This initiative aims to identify potential vulnerabilities and capabilities within frontier AI systems before they are deployed to the public. According to reporting from June 24, 2026, the current administration has initiated this request through confidential communications with the parent company of Facebook and WhatsApp, making Meta the last major US developer that has not yet formally joined the federal testing program. Meta confirmed it is currently in discussions with authorities regarding the proposed review process and intends to reach an agreement, citing support for the government’s goals of promoting secure and robust AI development.

Participation in these government reviews has become a standard industry practice among major AI developers. OpenAI and Anthropic already collaborate with federal agencies on testing unreleased models. Additionally, Google DeepMind, Microsoft, and xAI entered into agreements in May to provide officials with early access to their systems to mitigate national security risks. These safety measures are primarily intended to prevent the misuse of powerful AI models for cyberattacks, misinformation campaigns, or military operations, with particular concerns regarding autonomous identification of software vulnerabilities by models such as Anthropic’s Mythos AI.

The push for increased oversight follows recent government actions involving Anthropic, which was allegedly ordered to suspend access to its Fable 5 and Mythos 5 models for foreign nationals due to national security concerns. As Meta continues to negotiate the terms of its involvement, the administration maintains its stance that pre-release evaluations are essential for national security. While Meta has not officially finalized its participation, the company maintains that it remains committed to the development of safe and secure AI technologies within the United States.

The United States government is pressuring Meta to voluntarily submit its advanced artificial intelligence models for federal security reviews. This initiative aims to identify potential vulnerabilities and capabilities within frontier AI systems before they are deployed to the public. According to reporting from June 24, 2026, the current administration has initiated this request through confidential communications with the parent company of Facebook and WhatsApp, making Meta the last major US developer that has not yet formally joined the federal testing program. Meta confirmed it is currently in discussions with authorities regarding the proposed review process and intends to reach an agreement, citing support for the government’s goals of promoting secure and robust AI development.

Participation in these government reviews has become a standard industry practice among major AI developers. OpenAI and Anthropic already collaborate with federal agencies on testing unreleased models. Additionally, Google DeepMind, Microsoft, and xAI entered into agreements in May to provide officials with early access to their systems to mitigate national security risks. These safety measures are primarily intended to prevent the misuse of powerful AI models for cyberattacks, misinformation campaigns, or military operations, with particular concerns regarding autonomous identification of software vulnerabilities by models such as Anthropic’s Mythos AI.

The push for increased oversight follows recent government actions involving Anthropic, which was allegedly ordered to suspend access to its Fable 5 and Mythos 5 models for foreign nationals due to national security concerns. As Meta continues to negotiate the terms of its involvement, the administration maintains its stance that pre-release evaluations are essential for national security. While Meta has not officially finalized its participation, the company maintains that it remains committed to the development of safe and secure AI technologies within the United States.