NewsBite analysis · AI, Business, Regulation, SecurityThe Decoder

Trump's new executive order wants AI companies to voluntarily submit models for government safety reviews

The White House issued an executive order aimed at bolstering AI innovation and security by directing federal agencies to strengthen cyber defenses of government systems within 30 days. The Order tasks the Department of Defense, CISA, and the Treasury to establish a vulnerability-clearinghouse in partnership with the AI industry, signaling a push for coordinated risk management across government and industry. It also creates a voluntary framework for what it terms 'covered frontier models'—asking developers to submit these advanced systems to safety testing before release, but explicitly stopping short of a mandatory approval regime. Industry reactions are mixed. Some major players—Google DeepMind, Microsoft, and xAI—have already committed to pre-release model reviews through existing industry-government initiatives. OpenAI and Anthropic have advocated a tighter governance regime, arguing that democratic oversight and independent audits are essential for safety. The balance between safety and innovation remains a central question, with concerns that voluntary tests could tilt the playing field in favor of well-funded incumbents and restrict non-U.S. and open-source efforts. The policy comes amid ongoing U.S.-China tech competition and debates over how to regulate frontier AI without crippling progress. The EO clarifies that participation is voluntary and asserts the government will not impose excessive rules. But the push for a formal safety framework and faster cybersecurity hiring could accelerate spending on governance, testing, and compliance across firms of all sizes. In parallel, OpenAI's policy paper advocates binding national standards and whistleblower protections, signaling a broader debate about who writes the rules for frontier AI.