The U.S. government is currently pressuring Meta to submit its advanced AI models for a federal safety review before public release.
This situation places Meta in a difficult position, as it is now the only major American AI lab that hasn't agreed to this kind of pre-release evaluation. Five of its biggest competitors—Microsoft, Google, xAI, OpenAI, and Anthropic—have already established arrangements for government testing. The White House is using a June 2nd executive order, which created a voluntary 30-day review window, as its main tool of persuasion.
So, why is this happening now? The pressure campaign is built on a few key factors. First, the government's stance on AI has shifted dramatically towards prioritizing national security. A recent incident involving an AI model from Anthropic, named Mythos, which showed advanced capabilities in discovering cyber vulnerabilities, set off alarm bells in Washington. This event gave officials a powerful reason to push for an 'evaluate before you deploy' policy.
Second, peer pressure is playing a major role. When Microsoft, Google, and xAI joined OpenAI and Anthropic in May by agreeing to pre-release reviews, they created a new industry standard. This mass agreement makes Meta's reluctance stand out, inviting political scrutiny and making it look like an outlier. The government's request seems more reasonable when everyone else has already said yes.
Finally, the government made compliance easier. An earlier draft of the executive order proposed a lengthy 90-day review period, which the industry pushed back against. The final version was shortened to just 30 days. This change lowered the barrier to entry, making it a much more manageable request for busy tech companies and encouraging wider participation.
For investors, this standoff creates uncertainty. Since the executive order was signed, Meta's stock has fallen by nearly 6%, reflecting concerns about potential regulatory hurdles and increased costs. The company is already planning massive capital expenditures of up to $145 billion in 2026 for its AI infrastructure. Any forced delays or additional evaluation work could impact these plans and, consequently, its valuation.
- Frontier Model: A highly capable, general-purpose AI model that can perform a wide variety of tasks and could pose significant risks to public safety.
- Executive Order (EO): A directive issued by the President of the United States that manages operations of the federal government. It has the force of law.
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, or equipment.