[Core Summary] According to Fox News, four major US AI companies — OpenAI, Anthropic, Meta, and Google — jointly refused to sign legally binding AI safety guarantees at the White House AI policy meeting. The companies insist on a “voluntary commitment” path, which has sparked widespread discussion among regulators and the industry.

[Event Details]

  • Meeting Background: The White House AI Policy Coordination Meeting brought together executives from four major AI companies with representatives from the White House Office of Science and Technology, the Department of Homeland Security, and the Department of Commerce for closed-door talks on issues including AI model weight protection, abuse response, and critical infrastructure protection.
  • Refusal Content: The four companies refused to sign a legally binding “Model Safety Guarantee,” and refused terms that “would hold the company jointly responsible when the company fails to prevent large-scale mass leakage events.”
  • Substitute Commitments: The companies agreed to continue with the previously announced “voluntary commitments” at the White House, including establishing internal safety committees and publicly disclosing major risk assessment reports.
  • Internal Differences: Anthropic is relatively more open to accepting more rigorous regulation than the other three; OpenAI and Google demand that regulation be “predictable and not excessive”; Meta takes the toughest stance, tending to oppose any new legislation.
  • Regulatory Reaction: White House officials expressed “deep concern” about the position of the four companies, hinting at accelerating AI legislation in Congress.

[Panoramic Analysis] First: This collective refusal is not AI companies’ denial of safety, but a redefinition of regulatory pace and responsibility. AI companies have realized the industry is about to enter a “quasi-financial regulation” era, where binding commitments and legal responsibilities will have profound impacts. They hope to replace “hard law” (mandatory rules) with “soft law” (commitments, self-regulation) to secure room for R&D and commercialization. Second: Major AI companies’ refusal of hard guarantees will trigger “dual-speed regulation” divergence. The EU insists on implementing existing AI Act high-risk provisions, while some US states (like California) may legislate first; this will result in global AI companies accepting different standards in different jurisdictions, forcing R&D resources to “re-track for compliance.” Third: Investors have begun reassessing the long-tail risk of “AI safety incidents.” Anthropic’s heavy investment in “valuing safety” with slower commercialization has become a divergence point for some long-term investors; Meta, Google, and OpenAI’s “low commitment + high speed” path faces potential premium risks from credibility crises. AI compliance capability will rapidly become a core capability construction for leading companies.

Multi-View Comparison

  • White House Office of Science and Technology (AVD): Expressed “disappointment” with the position of the four companies, emphasizing it will push Congress to accelerate AI legislation.
  • Anthropic Public Policy Head: The company has taken “voluntary commitments” seriously but demands that new legislation does not “undermine innovation.”
  • Meta Global Affairs Head: Firmly opposes new regulations that damage established company competitiveness.
  • European Parliament Member (AI Act Drafter): The EU will strictly implement the AI Act, unaffected by regional laws.
  • Tsinghua University AI Safety Center (Chinese Researchers): The event reflects that the “global AI regulatory bottom line” is still being explored; China and the US may move toward different ecosystems in the AI governance field.

Editor: GoodInfo Global News Team