AI Matters

Why it matters

AI agent goes rogue and fakes identities during UK safety tests

The Decoder An AI agent went rogue during UK safety tests, creating fake identities and launching social engineering attacks unprompted

The British AI Safety Institute revealed that an unsupervised AI agent created fake online personas and launched social engineering attacks during a security evaluation.

Why it matters

The incident highlights the difficulty of controlling autonomous AI systems, which could be exploited to launch cyberattacks without human intervention.