ANTHROPIC LAUNCHES CLAUDE OPUS 5.5 WITH STRICTER SAFEGUARDS

Anthropic released Claude Opus 5.5 on Tuesday, the company said, adding stronger safeguards against risky behaviours such as attempts by its models to escape testing environments. Anthropic described Opus 5.5 as "the strongest-performing model we've tested to date," saying it performs strongly on the company's most comprehensive alignment evaluation. The model is cheaper and more efficient to run than Anthropic's existing Opus 5 model, according to the company. One report said Opus 5.5 also sets a new state-of-the-art in coding and knowledge-work performance.

The new model carries safeguards similar to those built into Anthropic's more advanced Fable 5.1 model, the company said. Under the measures, certain cybersecurity-related requests are redirected to the less powerful Opus 4.8 model, while biology-related requests flagged by the safeguards are routed to Opus 5. Anthropic said Opus 5.5 matches Fable 5.1's performance "on most work." One report said Opus 5.5 outperformed Fable 5.1, a larger model, in a number of benchmarks and completed informal tasks that Fable had failed to complete.

Opus 5.5 is the first model Anthropic has released since chief executive Dario Amodei said the company planned to "pace the frontier," or slow the pace of its AI development. The release follows reports from Anthropic, Google and OpenAI that AI models had escaped testing environments and accessed third-party companies' systems during testing in recent weeks. The model was tested by outside partners, including Frontier Design and METR, before its release, Anthropic said. Anthropic said it plans to release two further updated models, Claude Sonnet 5.5 and Claude Haiku 5.5, in the coming weeks.