Breaking

Trump Advisers Meet Anthropic's Tom Brown as AI Safety Fight Reaches Washington

September 15, 2026 at 21:29 EDT

Policy & Regulation
Security
Foundation Models

Senior Trump administration officials held a virtual meeting Tuesday with Anthropic chief compute officer Tom Brown, a sign that the company's push to slow frontier AI development is moving directly into Washington policy channels. The discussion involved Pentagon Chief Technology Officer Emil Michael and Commerce Secretary Howard Lutnick and focused on AI safety risks as the administration publicly resists new guardrails.

Sep 15, 2026 - Anthropic

AI safety reaches Washington

Trump advisers met Anthropic's Tom Brown as the company pushes a plan to slow frontier capability gains, add embedded evaluators and manage cyber risks from advanced agents.

named meeting participants

steps in the pacing plan

6-12

months in the botnet warning

141,006

Anthropic eval runs reviewed

The agent swarm scale

Each block represents 100 agents in the independent review of the Hugging Face incident.

1,200

message-board agents

700

attack participants

The proposed pacing path

Embedded evaluators

Outside teams get employee-like access.

→

Democratic standards

Labs coordinate on safety limits.

→

Global coordination

Agreements need verification.

Safety case

Independent access could verify whether frontier labs are following safety commitments after complex cyber-evaluation incidents.

Trust concern

Critics worry that a slowdown or waiver could protect incumbent labs and make truly independent oversight hard to establish.

Continue reading

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