Claude Opus 4.7 Flags Hantavirus Vaccine Questions as Safety Risk, Pausing Chats

A Hacker News user reported that asking Claude (Opus 4.7) "How would you develop a vaccine for the hanta virus?" triggers a safety filter, producing a modal that says: "Chat paused. Opus 4.7's safety filters flagged this chat. Due to its advanced capabilities, Opus 4.7 has additional safety measures that occasionally pause normal, safe chats. We're working to improve this. Continue your chat with Sonnet 4, send feedback, or learn more."

The user attempted to work around it by asking about the virus itself first, but the vaccine query still got flagged in a new session. Another user confirmed the same result with a more specific prompt: "How would you develop a vaccine for the hanta virus, specifically the Andes virus?"

Commenters highlight that public health researchers discuss vaccine design methods openly in papers every day, and that such aggressive blocking harms normal educational use. One user noted that the issue extends beyond Opus 4.7: when they created a group of expert personas to build predictive models for Hantavirus in the US (2025–2027), even Sonnet 4.6 returned the same chat paused error.

The broader discussion points to corporate liability as the driver—tech companies fear lawsuits if users misuse AI for harmful purposes, leading to broad over-filtering of sensitive topics like health and biosecurity.