Self-Replicating 'AI Worms' Could Soon Infect Connected Devices, Anthropic Cautions

Self-Replicating 'AI Worms' Could Soon Infect Connected Devices, Anthropic Cautions

AI research firm Anthropic has issued a warning that autonomous "AI worms" capable of self-replication could be just months away from spreading across internet-connected devices, including smart home appliances.

Leading artificial intelligence (AI) research company Anthropic has issued a cautionary alert regarding self-replicating autonomous AI systems that could begin proliferating across user devices independently, mirroring the behavior of traditional computer viruses.

According to a Monday report published by researchers William Zhou and Daniel Freeman from Anthropic's Frontier Red Team, malicious "AI worms" could soon begin autonomously infiltrating common household devices, such as Android TVs, smart home appliances, security cameras and various other internet-connected gadgets equipped with rudimentary computing capabilities and inadequate security measures.

Compact AI models could be "months away from being capable enough to self-replicate on commercial internet-of-things (IoT) devices," presenting a formidable security challenge as such a threat would possess "no centralized point to shut off," the report's authors noted.

This alert emerges amid heightened worries surrounding the expanding mathematical prowess of AI models, which could potentially compromise the cryptographic signatures that secure crypto wallets, putting approximately one-third of the Bitcoin supply at risk.

Nearly a third of the world's Bitcoin supply, equivalent to 6.26 million BTC, is currently stored in addresses that are susceptible to being compromised by AI models or quantum computers. Nevertheless, 4.33 million of these coins face exposure due to address reuse, indicating that transferring that BTC to new wallets would eliminate the vulnerability, according to Glassnode co-founder Rafael Schultze-Kraft in a Thursday X post.

Just one day prior, Ethereum researcher Justin Drake issued a warning that accelerating AI developments could compromise the elliptic curve digital signature algorithm (ECDSA) that protects cryptocurrency wallets earlier than previously anticipated and encouraged users to transition to fresh wallets.

Haseeb Qureshi, managing partner at Dragonfly, dismissed the warning as "cryptographic doomerism" and stated that transferring tokens to new addresses did not constitute a genuine solution to counter the threat.