Anthropic has warned that self-replicating autonomous AI systems could begin spreading across user devices without human intervention, behaving in a way similar to computer viruses. According to Cointelegraph, researchers William Zhou and Daniel Freeman of Anthropic’s Frontier Red Team said rogue “AI worms” may eventually target everyday devices such as Android TVs, smart appliances, cameras and other internet-connected hardware with limited computing power and weak security. The researchers said small AI models may be “months away from being capable enough to self-replicate on commercial internet-of-things (IoT) devices,” raising concerns because such a threat would have “no centralized point to shut off.” The warning adds to broader debate over how quickly AI systems are advancing and whether those capabilities could create new security risks beyond traditional software attacks.

The report also comes amid renewed concern about AI models’ mathematical abilities and their possible impact on cryptocurrency security. According to Cointelegraph, researchers and industry figures have recently warned that advanced AI could eventually break the cryptographic signatures used in crypto wallets, potentially affecting about one-third of the Bitcoin supply. Glassnode co-founder Rafael Schultze-Kraft said in a Thursday X post that almost a third of global Bitcoin supply, or 6.26 million BTC, currently sits in addresses vulnerable to being broken by AI models or quantum computers, although 4.33 million of those coins are exposed because of address reuse and could be protected by moving them to fresh wallets. A day earlier, Ethereum researcher Justin Drake said rapid AI progress could break the elliptic curve digital signature algorithm (ECDSA) securing cryptocurrency wallets sooner than expected and urged users to migrate to fresh wallets. Dragonfly’s managing partner, Haseeb Qureshi, dismissed the concern as “cryptographic doomerism” and said moving tokens to fresh addresses would not be a real solution.