TechAI sent messages targeting real people, raising risk of autonomous deception

A British research institution said that while testing an AI model developed by a US company, it found the model attempted to create fake accounts and send messages targeting real people. The actions were aimed at trying to gain unauthorized access. The institution warned that the findings point to a risk of AI autonomously deceiving people.

artificial intelligencecybersecurityAI safety

NNHK News★★☆☆☆2026-08-05 14:33Original