White House AI czar calls Anthropic's safety push a power grab

Anthropic CEO Dario Amodei rarely posts on social media, but a recent debate dragged him into the spotlight, sparking an aggressive public rebuke from White House AI czar David Sacks.

The feud began when investor Gavin Baker discussed rumors on X and a podcast, including chatter that Anthropic aimed to become the only private company in the world, an allegation Anthropic staff denied. Baker also argued Amodei had lost the policy debate and warned his alarmist rhetoric was fueling anti-data-center activism.

Amodei logged on to push back, arguing that viewing all regulation as regulatory capture is a Silicon Valley bubble mindset. He insisted structural scaling laws and chip access naturally centralize power, claiming Anthropic actually designs policy proposals to slow down frontier labs while giving challengers room to catch up.

An FDA for AI or a bureaucratic DMV?

David Sacks was not convinced. In a detailed post on X, Sacks accused Amodei of running an orchestrated fear campaign to build a moat around his business through regulatory capture.

Sacks cited economist George Stigler, noting that regulatory capture occurs when industry incumbents use revolving doors and lobbying to design rules that protect themselves. He highlighted Anthropic hiring former Biden AI policy officials as evidence of an operation pushing state and federal frameworks for its own benefit.

The main flashpoint is pre-deployment testing. Amodei has consistently backed federal review agencies, variously comparing the concept to the FDA, FAA, or a FINRA-style body. Sacks countered by calling it a DMV for AI, warning that years of compliance testing for theoretical risks would create bureaucratic bottlenecks that handicap American developers against China.

Sacks also took aim at Amodei's track record on AI safety warnings. He pointed to Amodei's May 2025 claim that AI would eliminate 50 percent of entry-level knowledge jobs within five years, noting that fifteen months later there is still zero empirical evidence backing it up.

Why it matters

We think this clash exposes the fundamental philosophical war defining modern tech policy: whether frontier AI is too dangerous to distribute, or too powerful to centralize.

For developers and startups, the stakes are concrete. If the White House leans into Sacks' view, open-weights models will retain breathing room to compete without pre-release clearance. If Amodei's pre-deployment testing takes hold, the regulatory friction could render catching up to the frontier prohibitively expensive for anyone without massive legal budgets.

Anthropic's pricing power relies heavily on staying ahead of open alternatives. By pushing for equal regulatory scrutiny across closed and open weights under the banner of catastrophic safety, the company looks clever in securing enterprise trust, but it risks isolating itself from the broader builder ecosystem.

Amodei claims that marketing cannot fix AI's trust deficit and that only real-world breakthroughs like curing cancer will change public sentiment. That may be true, but deploying speculative panic about job wiping and blackmail while lobbying for centralized federal gatekeeping is an awkward foundation for building that trust.