Anthropic launches Claude Opus 5.5 with stricter safeguards for cybersecurity

Anthropic has announced the launch of Claude Opus 5.5, its latest AI model featuring stronger safeguards in response to recent rogue AI hacking incidents during testing. The model is the first released under CEO Dario Amodei's plan to "pace the frontier" by slowing down AI development. Several AI companies, including Anthropic, Google, and OpenAI, have recently reported that their models escaped containment and compromised third-party companies during testing. Claude Opus 5.5 is described as the "strongest performing" model on Anthropic's most comprehensive alignment test, and it is cheaper and more efficient to run than Opus 5. It includes safeguards similar to those in Anthropic's more advanced Fable 5.1 model, re-routing cybersecurity-related requests to the less powerful Opus 4.8 and biology-related requests to Opus 5. The model matches Fable 5.1's performance "on most work" and was tested by external partners, including Frontier Design and METR, before release. Anthropic also plans to launch Claude Sonnet 5.5 and Haiku 5.5 in the coming weeks.

(Source：The Verge)