The U.S. government, which imposed export controls on Anthropic's advanced artificial intelligence (AI) model, is pressuring Meta to submit its new AI model for a government safety review.
On the 23rd (local time), the New York Times (NYT) reported, citing anonymous sources, that the U.S. government has been exchanging confidential emails with Meta as part of this pressure.
Meta is the only one among major AI developers in the United States that has not signed a safety review agreement with the government. OpenAI, Anthropic, Google, xAI and Microsoft have all submitted their new AI models to the government's AI Standards and Innovation Center (CAISI), the AI safety review body, for evaluation.
In April, Meta unveiled its latest model, "Muse Spark," which was assessed to have performance nearly on par with major competitors' models.
Francis Brennan, Meta Spokesperson, said, "We share the government's goal of cementing U.S. leadership in powerful and safe state‑of‑the‑art AI. We are currently working through specific terms and hope to sign the agreement soon."
Ben Cas, Commerce Department Spokesperson, also said the CAISI under the department is regularly consulting with corporations on the agreement, explaining, "This (email exchange) is routine. That is exactly what CAISI does."
Two weeks earlier, citing national security concerns, the U.S. government ordered Anthropic to block foreign access to its latest models "Fable 5" and "Mythos 5." The government and Anthropic then held talks on the issue. President Donald Trump said in a recent interview with Axios that he no longer views Anthropic as a security threat.
On the 2nd, President Trump signed an executive order requiring a review of security issues in new AI models, setting up the government's safety review process. Under the order, technology corporations must allow the U.S. government to evaluate new AI models for up to 30 days before release. During this period, the government can assess the AI's capabilities and vulnerabilities.