Breaking

David Sacks Says Anthropic and OpenAI Can Slow AI Work Without Washington’s Intervention

September 14, 2026 at 21:05 EDT

Policy & Regulation
Foundation Models
AI Agents

David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, said Anthropic and OpenAI should make their own product and release decisions if they believe frontier AI progress needs to slow, rather than seeking government intervention. The remarks put the Trump administration’s light-touch AI posture in direct tension with a new push from leading AI executives for coordinated safety controls.

Sep. 2026 · AI policy

AI labs are told to set their own pace

David Sacks says Anthropic and OpenAI can slow frontier AI development themselves if they believe the risk warrants it. The debate now turns on whether voluntary action can substitute for government-enabled coordination.

steps in Amodei's pacing plan

6-12

months in the cited risk window

1-2

years to advance alignment and testing

3-5

years cited as geopolitically important

The cited risk window, drawn to scale

6 months

lower bound

12 months

upper bound

Each block represents one month in the window Amodei described for a more capable agent swarm risk.

What pacing would require

1. Embedded evaluators

Outside teams inspect safety practices and incidents with employee-like access.

→

2. Democratic standards

Frontier companies align on safety thresholds and limits within democratic countries.

→

3. Global coordination

Governments try to coordinate on the hardest cross-border risks.

Supporters' case

External evaluators and shared standards could make safety commitments more verifiable before the most capable models are released.

Critics' concern

Large labs may be able to slow rivals or shape rules in their favor if pacing becomes industry coordination.

Continue reading

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