Breaking

Anthropic releases Claude Opus 5.5 days after Dario Amodei urges slower AI progress

September 22, 2026 at 13:54 EDT

Foundation Models
AI Agents
Policy & Regulation

Anthropic released Claude Opus 5.5 on September 22, days after CEO Dario Amodei argued that frontier AI labs should slow capability gains so safety work can catch up. The launch sharpens a live industry debate: Anthropic is calling for “pacing the frontier” while continuing to ship more capable models and expand the internal use of AI agents in its own research.

Sep. 22, 2026 · Anthropic

Pacing the frontier, shipping the frontier

Claude Opus 5.5 arrived days after Dario Amodei called for slower capability gains, putting Anthropic’s safety argument next to a major model release.

40%

Lower typical workload cost than Opus 5

26%

AI R&D work Claude now leads

30,000

Internal agents active at once

Cost shift from Opus 5 to Opus 5.5

Opus 5

100%

Opus 5.5

60%

Anthropic says typical workloads cost 40% less to run.

The proposed safety loop

Build models→Embed evaluators→Monitor agents→Release with safeguards

Why users may welcome it

Lower cost, faster output, stronger coding results, and clearer writing could make long agentic workflows easier to supervise.

Why critics are uneasy

A call to slow capability gains now sits beside a more capable model and large-scale internal AI-agent use.

Continue reading

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