Staff Reports
Demis Hassabis, the CEO of Google DeepMind, has published a manifesto and told reporters he wants the United States to create a new, industry-funded AI watchdog to vet the most powerful models before they launch. On its face, the idea sounds sensible: test what could be dangerous and stop runaway risks. But when Google’s top AI executive outlines a plan to build a regulator that industry helps fund and staff, conservatives should ask some hard questions about who really gets the power.
What Hassabis is actually proposing
Hassabis wants a FINRA-style body for “frontier” AI. Companies would submit models up to 30 days before public release so the watchdog can test for cyberattacks, biological threats, and deceptive behavior. He wants the board packed with top technical experts—Turing Award winners—plus industry, government, and open-source reps. At first the checks would be voluntary, but the plan moves to mandatory approval once the system works. Hassabis says the watchdog should cover models from anywhere in the world and could start before the end of the year. He’s been briefing President Trump’s officials and other leaders to build support.
Why this raises red flags for conservatives
Oversight is one thing. Letting Big Tech shape the rules, fund the gatekeeper, and help pick the experts who run it is another. That is the textbook recipe for regulatory capture: the industry writes the rules that keep rivals out and protect incumbents. Imagine a system where a “prestige badge” decides who can ship an AI model in the U.S. That badge could be used to freeze competition, favor certain political viewpoints, or give huge companies a chokehold on new tech. Conservatives should be skeptical when the very firms that profit most from AI promise to police themselves.
Security needs rules — but not Silicon Valley control
There was a real problem with the ad-hoc freeze on Anthropic’s Mythos and the scramble that followed. Washington needs clear, enforceable safety rules and a way to act during emergencies. But Congress—not private labs—should write the law. Any oversight body must have strict limits, transparency, congressional oversight, and sunset clauses so power doesn’t calcify in the hands of tech elites. Otherwise we trade one emergency scramble for a permanent licensing system run by the same people who got us here.
Bottom line
Yes, AI can be dangerous and it needs rules. No, we should not hand global gatekeeping to an industry-funded body that answers to Silicon Valley and unnamed “experts.” President Trump and Congress should welcome sensible safety standards, but insist on public control, clear legal limits, and competition protections. We need a watchdog, not a Silicon Valley lapdog with a shiny Turing Award collar.
