Breaking

Anthropic CEO Calls for Slower Frontier AI Development as Rivals Back Safety Plan

September 13, 2026 at 17:35 EDT

Foundation Models
AI Agents
Security

Anthropic CEO Dario Amodei is calling on frontier AI developers to slow the growth of model capabilities, arguing that alignment and security work need time to catch up. Leaders at OpenAI, xAI and Google DeepMind backed the direction within hours, but the proposal remains a voluntary, partly aspirational framework rather than a binding moratorium.

SEP 12, 2026 · ANTHROPIC

Frontier AI Calls for More Safety Time

Dario Amodei says capability gains are outpacing safeguards. Leaders at OpenAI, xAI and Google DeepMind backed the direction, but no industry-wide pause is in force.

The risk window Amodei highlighted

A possible gap between faster capability gains and slower safety work.

6 months

Lower bound

12 months

Upper bound

Takeaway: The warning concerns a possible capability horizon, not a confirmed present ability.

Proposed safety steps

141,006

Claude evaluation runs reviewed

Incidents identified

Pacing the frontier

Embedded evaluators

→

Shared safety bars

→

Global coordination

Why it appeals

More time for alignment and interpretability
Independent checks on training and deployment
Better operational security for agent evaluations

What remains unresolved

No binding industry-wide pause
Different views on what slower means
Geopolitical and enforcement risks

Continue reading

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