Trump Administration Pushes Meta to Submit AI Models for Voluntary Safety Assessment, Focusing on Security Vulnerabilities and Technical Capabilities
Meta is the only major U.S. AI company that has not yet reached an agreement with the U.S. government, while Microsoft, Google, OpenAI, Anthropic, and xAI have all agreed to submit their models for review by the Department of Commerce's AI Standards and Innovation Center. Meta spokesperson Francis Brennan confirmed that negotiations on the agreement details are ongoing, with hopes to sign soon.
This requirement stems from an executive order signed by Trump on June 2, which mandates that tech companies provide a maximum 30-day assessment period before releasing new models and requires a specific review process to be established by the end of July. The assessment aims to strengthen national security while maintaining a voluntary nature.
Source: Public Information
ABAB AI Insight
Meta has previously maintained a relatively independent stance in AI safety, publicly opposing certain mandatory regulatory measures and prioritizing the open-source Llama series, while competitors like OpenAI and Anthropic have repeatedly participated in government safety testing and quickly responded to takedown requests. Being the last major company to engage highlights its lag in policy coordination.
The Trump administration is mobilizing agencies like NSA and CISA through the executive order to establish a voluntary framework, encouraging companies to share model access rights in exchange for policy support and critical infrastructure protection cooperation. Meta's negotiation to join will further concentrate industry resources at the government-designated CAISI center.
Similar to the voluntary AI commitment agreements during the Biden administration, Google and Microsoft have also strengthened their ties with the government through similar mechanisms. Meta is currently in a transitional phase, trying to catch up with industry compliance while balancing its open-source strategy and regulatory pressures.
Essentially, this represents a regulatory change: the executive order quickly establishes an AI capability assessment mechanism in a voluntary form to avoid delays from formal legislation, while the 30-day pre-review window allows for early government intervention in cutting-edge models, accelerating capital concentration towards compliance-friendly companies and reshaping the interface between AI development and national security.
ABAB News · Cognitive Law
The longer the regulatory vacuum lasts, the higher the cost of policy catch-up.
Voluntary agreements are a soft landing; those who refuse will ultimately face a hard landing.
Technological leadership does not equate to policy safety; compliance is the new moat.