Earlier this month, Anthropic revealed it developed an AI system named Mythos Preview, which it chose not to release publicly out of caution. The company claims this AI can detect and exploit software weaknesses, raising alarms about potential dangers to economies, public safety, and national security. However, some experts question the actual extent of Mythos’ abilities. Ian Sample speaks with Guardian AI reporter Aisha Down to explore what Anthropic’s cautious approach indicates about its broader tactics and whether this move might prompt tighter industry oversight.
Back
