AI Worms Could Pose Threat to Smart Home Devices
Research by Anthropic indicates that internet-connected smart home devices could be at risk due to AI-powered malware.
AI company Anthropic has announced that malware that spreads by self-replicating across internet-connected devices could create a cybersecurity threat in the future. Researchers predict that small AI models could acquire this capability by the end of the year.
AI Worm Warning
In research published on October 12, Anthropic pointed out that low-cost, internet-connected devices could be used for self-replicating malware. It was emphasized that such software could continue to spread without human intervention.
Experiment on Android TV Devices
Researchers successfully deployed and ran an AI model on 15 cost-effective Android TV boxes. Although the replication process of the models was examined in a test network, no independent malware was observed emerging in the real world.
Evolution of Smaller Models
Certain models with 30 billion parameters or fewer are capable of performing a significant portion of the operations required for self-replication. Experts predict that these models could reach a level where they can fulfill all stages by the end of the year.
Increased Difficulty of Centralized Intervention
Unlike traditional malware, AI worms are said to be able to adapt to different devices. It is stated that disabling networks spread across a large number of devices from a single center could become difficult.
Expert Opinions and Recommendations
Georgetown University researcher Jessica Ji stated that the scenario is not out of the question, though the probability of it occurring is not yet known. Experts underline that security vulnerabilities must be closed and precautions increased.