Artificial intelligence company Anthropic has warned that self-replicating autonomous AI systems may begin spreading across user devices without human intervention, functioning similarly to computer viruses.
According to a report by Anthropic's Frontier Red Team researchers William Zhou and Daniel Freeman, rogue "AI worms" could autonomously target everyday devices including Android TVs, smart appliances, cameras, and other internet-connected devices with basic computing power and weak security.
The researchers noted that small AI models may be months away from possessing the capability to self-replicate on commercial internet-of-things (IoT) devices. Such a threat would prove challenging to address due to the absence of a centralized point that could be shut down.
The warning coincides with growing concerns about the mathematical capabilities of increasingly advanced AI models, which could potentially compromise the cryptographic signatures used in cryptocurrency wallets.
According to data cited in the report, approximately one-third of global Bitcoin supply, or 6.26 million BTC, currently resides in addresses considered vulnerable to being compromised by AI models or quantum computers. However, 4.33 million of those coins face exposure due to address reuse, a condition that could be resolved by moving the Bitcoin to newly generated wallets.
Ethereum researcher Justin Drake previously warned that rapid AI advancements could break the elliptic curve digital signature algorithm (ECDSA) securing cryptocurrency wallets sooner than anticipated, urging users to migrate to fresh wallets.
Not all observers share the same level of concern. Dragonfly's managing partner Haseeb Qureshi characterized the warning as "cryptographic doomerism" and disputed the effectiveness of migrating tokens to fresh addresses as a solution to the threat.
