Breaking

Anthropic CEO Dario Amodei Says He Would Accept Joint Democratic Oversight of Advanced AI

September 13, 2026 at 12:28 EDT

Policy & Regulation
Foundation Models

Anthropic chief executive Dario Amodei said he would accept placing the company's advanced AI under some form of joint oversight by "the right combination of governments," while rejecting a simple transfer of the technology to any single state.

September 13, 2026 · Anthropic

Joint governance, not a handover

Dario Amodei said advanced AI should sit under oversight by a combination of democratically elected governments. He warned that a single state could abuse the technology as easily as a single company.

The line that matters

"I don't know about handing over, but some kind of oversight, some kind of joint governance."

Who

A mix of democratically elected governments, not one capital

What

Oversight and shared rules, not a simple transfer of the company

When

A direction he said would be the work of years

The swarm warning, drawn to scale

Amodei estimated that an unaligned AI swarm could take over the internet in 6 to 12 months if capabilities keep racing ahead.

6 mo

Low end

12 mo

High end

Takeaway: the warning window is months, while the governance model he described would take years.

How he wants the frontier paced

Embedded evaluators

Independent inspectors with employee-like access. Anthropic is committing now.

→

Democratic coordination

Frontier labs in democracies set shared safety standards and pace limits.

→

Global coordination

Democracies try to bring in authoritarian states, with hard verification.

Support

Pace the frontier, with inspectors inside the labs

Rival lab leaders backed slower capability growth and independent evaluators. The first step is already a unilateral Anthropic commitment.

Pushback

Don't demand new rules as the price of slowing down

Critics say labs can slow themselves without a regulatory bargain, and that a single government could abuse the same systems. China matching any slowdown remains the hardest gap.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.