Three major American technology companies - Google, Microsoft, and xAI - have agreed with the White House to provide early versions of their AI models to the US government before public release, for security review and assessment.
According to the Wall Street Journal, this agreement represents the latest step in the US government’s efforts to strengthen AI safety regulation. The White House is considering establishing an AI model review mechanism that would require major AI companies to submit test versions of new models before public launch, allowing the government to evaluate potential safety risks.
The core provisions of the agreement include: tech companies must provide access to designated government agencies when model training is nearing completion; the government will conduct capability assessments and safety testing, with particular focus on whether models could potentially be used to create biological weapons, conduct cyberattacks, or other malicious purposes.
Industry analysts point out that this approach could become a benchmark for AI regulation in the United States. If more companies join, it would substantially change how AI models are developed and released. However, it also raises questions about whether government review could stifle innovation.
The European Union and the United Kingdom are also considering similar AI safety regulatory frameworks. Global coordination on AI regulation may be accelerating.