Anthropic has warned that self-replicating artificial-intelligence systems could soon spread between internet-connected devices without human intervention, creating a threat that would resemble the behavior of a computer virus.
AI systems could target everyday devices
According to Cointelegraph, researchers William Zhou and Daniel Freeman of Anthropic’s Frontier Red Team said rogue “AI worms” could eventually target devices such as Android televisions, smart appliances, cameras and other internet-of-things equipment. These products often have basic computing capabilities and may have weaker security protections than conventional computers.
The researchers said smaller AI models could be only months away from becoming capable enough to reproduce themselves on commercial IoT devices. If that happens, an infected system could potentially use its capabilities to seek out and compromise other connected equipment, rather than relying on a person to manually spread malicious software.
Anthropic’s warning concerns the possibility of autonomous replication: an AI system copying or extending itself across multiple devices. The researchers described the prospect as particularly difficult because such a network would have “no centralized point to shut off.” In practical terms, that could make containment more complicated than disabling a single server or service.
Why the warning matters for crypto security
The alert comes as researchers and industry figures are also debating whether advances in AI could threaten the cryptographic systems used to protect cryptocurrency wallets. Cryptographic signatures are mathematical mechanisms that help prove ownership and authorize transactions.
Glassnode co-founder Rafael Schultze-Kraft said that 6.26 million Bitcoin, or almost one-third of the global supply, is currently held in addresses potentially vulnerable to attacks by AI models or quantum computers. He said 4.33 million BTC of that amount is exposed because of address reuse, meaning that transferring those coins to fresh wallets would remove that particular exposure.
Ethereum researcher Justin Drake separately warned that progress in AI could undermine the elliptic curve digital signature algorithm, known as ECDSA, sooner than previously expected. ECDSA is used to secure cryptocurrency wallets. Drake urged users to move assets to fresh wallets, although Dragonfly managing partner Haseeb Qureshi dismissed the warning as “cryptographic doomerism” and said moving tokens to new addresses was not a complete answer to the underlying threat.
These concerns are related to the broader question of whether increasingly capable AI systems could discover weaknesses, automate attacks and operate at a scale that is difficult for people to control. Anthropic’s report focuses on connected consumer devices, while the crypto discussion centers on the mathematical protections behind digital assets.
Moneyiar market picture
Moneyiar data at 20:34 on October 13, 2026, showed all four tracked assets lower over both 24 hours and seven days:
- Cronos (CRO): $0.0588544373, down 4.51% in 24 hours and 11.42% over seven days.
- Lighter (LIT): $3.366604023, down 8.04% and 12.69%, respectively.
- Monad (MON): $0.0244671984, down 1.96% and 14.96%.
- ONEchain (ONE): $0.1183169839, down 5.15% and 17.99%.
The market figures describe current price performance and do not establish a connection between those moves and Anthropic’s cybersecurity warning.
