Opus 5.5 refuses basic molecular biology questions (bede.im)

Opus 5.5, Anthropic's latest language model, has been criticized for its overly cautious biosecurity protocols, which led it to refuse to answer fundamental molecular biology questions that were adequately handled by other open-weight models. In a comparative analysis, the model rejected a query regarding the use of Saponin-DNase in selective lysis, citing biosecurity concerns, while competitors like Kimi K3 and DeepSeek V4 Pro both readily provided comprehensive responses. The disparity highlights a significant limitation for researchers relying on Opus 5.5, particularly when the model's self-imposed restrictions hinder its utility in scientific exploration. The findings underscore the ongoing tension within the AI community between ensuring safety and providing access to valuable scientific information. Despite Opus 5.5's advanced capabilities, it has proven less reliable for research purposes than its open-weight counterparts, which demonstrated not only the ability to answer complex questions but also to synthesize information effectively. As researchers encounter increasing challenges with safety protocols in AI models, the utility of other models in conducting molecular biology research may drive a shift towards more accessible alternatives, emphasizing the need for a balanced approach in AI development that fosters both innovation and safety.