Breaking

Trump Adviser David Sacks Tells AI Labs to Prove Safety Themselves Amid Slowdown Push

September 14, 2026 at 23:01 EDT

Policy & Regulation
AI Agents
Security

David Sacks, a top technology adviser to President Donald Trump, has challenged calls for an AI slowdown by putting first responsibility for safety on the companies building frontier models. The intervention, published September 14, escalates a fight over whether labs should voluntarily slow deployment, submit to outside evaluators, or continue racing to preserve a U.S. lead over China.

September 2026 · Frontier AI

AI Safety Moves Back to the Labs

David Sacks says frontier developers must prove their systems are safe, while Dario Amodei is pushing a slower pace for the most capable models.

6-12

months in Amodei's agent-swarm warning

parts in the pacing proposal

harm areas in Anthropic's report

Risk window visualized

Amodei flagged a six to 12 month window for a more capable misaligned agent swarm.

6 months

earliest window

12 months

upper window

Proposed pacing path

Embedded evaluators

Ongoing employee-like access to verify safety practices.

→

Shared standards

Democratic labs coordinate safety thresholds and release limits.

→

Global talks

Governments test whether any wider agreement can be verified.

Support case

Outside evaluators could make safety claims more verifiable and expose incidents sooner.

Skeptic case

Voluntary pacing could protect incumbents, burden open-source rivals or replace enforceable limits with paperwork.

Sacks counterpoint

Developers should make products safe themselves; if they cannot control a system, they should not release it.

Lab responsibilityNo broad U.S. pauseProceed safely

Continue reading

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