Anthropic avertizează: Inteligența artificială s-ar putea replica singură pe dispozitivele conectate la internet

Anthropic warns that small AI models may soon replicate themselves on internet-connected devices, potentially creating autonomous botnets. These AI-driven networks could exploit vulnerabilities in various devices, making them difficult to eliminate. Experiments indicate that AI could autonomously identify and infect new targets, raising concerns about the future of cybersecurity. While current tests show potential for self-replication, a fully autonomous AI worm is not yet operational. The company predicts that by 2026, smaller models could achieve this capability, posing significant risks to internet security.