Breaking

OpenAI Confirms Weeks of AI Safety Coordination With Anthropic and Google DeepMind

September 15, 2026 at 11:49 EDT

Policy & Regulation
Foundation Models

OpenAI has been coordinating artificial intelligence safety work with rivals Anthropic and Google DeepMind for several weeks, the company's global policy chief said Tuesday, even as President Donald Trump's administration dismissed existential-risk warnings as a hoax and pressed U.S. labs to stay ahead of China.

September 15, 2026 · OpenAI

Rival labs open a safety channel

OpenAI has been coordinating AI safety with Anthropic and Google DeepMind for several weeks. The White House, meanwhile, has dismissed existential-risk warnings as a hoax and urged the United States to stay ahead of China.

Risk window

A misaligned swarm, 6 to 12 months

Amodei warned that, at the current pace, a more capable misaligned agent swarm could take over the internet within 6 to 12 months.

6 mo

Lower bound

12 mo

Upper bound

Takeaway: The taller column is twice the shorter one — Amodei’s outer bound is 2× the inner bound.

Frontier labs in the talks: OpenAI, Anthropic, Google DeepMind

Weeks

Lehane said coordination has already been under way for several weeks

1–2 yr

Extra time Amodei said could greatly reduce the chance something goes seriously wrong

Three-step plan

How Amodei would pace the frontier

STEP 1

Embedded evaluators

Employee-like third-party access. Anthropic is committing now; OpenAI said it will follow.

→

STEP 2

Democratic coordination

Shared standards and limits on unchecked progress among labs in democratic countries.

→

STEP 3

Global coordination

Harder agreements, including with authoritarian governments, with verification limits.

The case for talking

Safety-focused researchers and some developers say the leading labs should set shared guardrails rather than race in isolation, comparing safety cooperation to commercial aviation.

The case against

The White House calls the alarm a hoax that would help China. Critics warn the talks could look like a cartel or regulatory capture by the largest labs. No binding pact has been announced.

The political split

Industry wants time and independent reviewers. The White House wants speed.

LEHANE

"It's better to try to work together to prioritize safety."

TRUMP

"Whoever wins AI, wins" — safety fears dismissed as a hoax that would please China.

Continue reading

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