Anthropic Warns AI Worms Could Self-Spread Across Devices, Threaten Bitcoin Wallets

Anthropic has warned that self-replicating AI systems could autonomously spread across users’ devices like computer viruses.

Cointelegraph reported on October 13 that Anthropic Frontier Red Team researchers William Zhou and Daniel Freeman wrote in a report that small AI models could reach a level within months where they are capable of self-replication on commercial internet-of-things, or IoT, devices.

The researchers said such AI worms could target internet-connected devices with weak security, including Android TVs, smart home appliances and cameras. That would make them difficult to contain because there is no central point at which they can be blocked.

Concerns are also rising that more capable AI could threaten the security of cryptocurrency wallets. Glassnode co-founder Rafael Schultze-Kraft wrote on X, formerly Twitter, that 6.26 million Bitcoin, or about one-third of the token’s total supply, is stored at addresses that could be hacked by AI models or quantum computers.