Sacks and Wiles Shelve White House FDA-for-AI Plan
Key insights
- NEC Director Hassett's FDA-for-AI proposal was floated and walked back by Wiles and Sacks within the same news cycle.
- AI czar David Sacks has reframed AI governance away from domestic safety toward Chinese AI competition as the primary threat.
- CAIS quietly removed its frontier AI testing webpage, signaling institutional deprioritization of domestic model safety review infrastructure.
Why this matters
The public collapse of an AI safety proposal inside the White House signals that the U.S. government lacks a coherent framework for governing frontier AI, which directly affects how labs like Anthropic and OpenAI will face regulation going forward. Sacks's pivot to framing China as the primary threat suggests future executive actions will favor export controls and national security framing over domestic safety reviews, reshaping how AI policy advocacy needs to be positioned across the industry. The removal of CAIS's testing infrastructure means the technical scaffolding for independent frontier model evaluation is losing institutional backing at exactly the moment when labs are releasing their most capable systems.
Summary
The Trump administration's AI safety push collapsed into internal contradiction this week. NEC Director Kevin Hassett floated a formal 'FDA for AI' model for frontier AI systems. Chief of Staff Susie Wiles and AI czar David Sacks shut it down within hours, reframing the core threat as Chinese AI capabilities, not domestic lab safety.
The Center for AI Standards and Innovation quietly removed its frontier AI testing webpage during the fallout, suggesting the institutional scaffolding for any FDA-style review process is already being wound down.
Essentially: (Hassett vs. Sacks/Wiles) are split over whether AI governance means regulating domestic labs or outcompeting Beijing.
- Momentum for the executive action had spiked after Anthropic's Mythos release but industry sources say it has since largely dissipated.
- CAIS's webpage removal signals that federal infrastructure for independent frontier model evaluation is losing support at the same moment it was most needed.
The White House has no unified theory of AI risk, leaving frontier model governance in limbo through the rest of 2026.
Potential risks and opportunities
Risks
- Without a governing framework, frontier labs like Anthropic and Google DeepMind face regulatory uncertainty that complicates enterprise contracts and capital raises requiring compliance disclosures.
- If Sacks's competition framing dominates policy, domestic AI safety research organizations such as METR and Redwood Research may lose federal funding and policy influence within the next 12 months.
- The CAIS testing webpage removal could leave no official U.S. evaluation channel for frontier models, creating a governance gap that EU AI Act enforcers may exploit to impose external standards on American labs.
Opportunities
- Domestic AI safety evaluation firms like METR and Apollo Research can position as the de facto frontier testing layer if CAIS vacates the space, attracting direct lab contracts.
- Export control consultancies and national security-focused AI firms such as Palantir and Shield AI gain policy relevance as Sacks's China-threat framing elevates their standing in the executive action drafting process.
- Think tanks aligned with the competition-over-regulation framing, including CNAS and Hudson Institute, have an opening to shape the executive action's final text while Hassett's faction remains sidelined.
What we don't know yet
- Whether Anthropic lobbied for or against the FDA-style review model after Mythos's release triggered renewed White House momentum
- What specifically prompted CAIS to remove its frontier AI testing webpage, and whether that action was coordinated with Sacks's office
- Whether Sacks's 'Chinese AI threat' reframe will be codified into any eventual executive action, or whether the internal split means no action proceeds at all in 2026
Originally reported by axios.com
Read the original article →Original headline: White House AI Executive Action Stalls as NEC Director's 'FDA for AI' Proposal Is Immediately Walked Back by Chief of Staff Wiles and AI Czar Sacks