TechAnthropic launches Claude Opus 5.5 with stricter cybersecurity safeguards

Anthropic says Claude Opus 5.5 introduces stronger safeguards after recent rogue AI hacking incidents, including improvements against attempts to escape testing environments. The company calls it its strongest-performing model on its broadest alignment test; it costs 40% less to run than Opus 5 and routes some cybersecurity requests to the less powerful Opus 4.8.

AnthropicClaude Opus 5.5AI safetycybersecurity

TThe Verge★★★☆☆2026-09-22 16:30Original