Anthropic report describes a self-replicating AI worm

Anthropic's offensive research team published a report this week describing a practical, self-replicating AI worm, according to The Guardian on October 18, 2026. The report's central claim is that such worms would need more capability to work effectively on less powerful devices, and that current trends make this feasible by the end of the year. The Guardian noted that warnings about what AI might do in the future, along with alarming reports of what it has already done, are being issued so quickly that even security experts have trouble triaging what to worry about. What drew attention to Anthropic's report was that it described a new, practical worm. Public information currently stops at the report's release; the materials do not disclose the worm's name, the test environment, or further technical details.