Anthropic blocks possible attempt to use AI to make biological weapons

JUST IN: Anthropic says it blocked a possible attempt to use its AI to help make biological weapons, revealing the case in a new threat intelligence report as experts urge stricter access to advanced models.

Anthropic says people tried to use its AI to help build biological weapons.

The company laid it out in a new threat intelligence report, five case studies of actors using Claude in ways that could support bioweapons development, all blocked between December and August. And it wasn't alone. Google said this week someone asked Gemini for a complete step-by-step guide to synthesising weaponised biological agents. Anthropic's list goes further, a Russia-linked espionage campaign, an Iranian propaganda outfit, surveillance built to identify dissidents, and Chinese firms allegedly trying to clone Claude itself.

Washington is split on what to do. Bernie Sanders wants a pause on advanced AI and an outright ban on superintelligence, you can see him making that case here. President Trump is having none of it, saying his worry is that if America doesn't win AI, it ends up in a very bad position.

Anthropic's own words are the ones that stay with me. Without the right safeguards, these capabilities could have catastrophic consequences.

This is Liu Yang, reporting for HIT.