Anthropic’s Frontier Red Team has issued an alert that autonomous artificial-intelligence programs, dubbed “AI worms,” could soon spread across consumer devices without human direction, resembling traditional computer viruses.

AI worms and IoT devices

The team, led by William Zhou and Daniel Freeman, notes that modest-sized AI models may be capable of self-replication on internet-connected appliances such as smart TVs, cameras, and other low-power gadgets within a few months. Because many of these products lack robust security measures, a worm could propagate without a central control point, making containment difficult.

Potential impact on cryptographic security

The warning arrives as concerns grow that increasingly powerful AI could undermine the cryptographic signatures protecting digital assets. Analysts estimate that roughly one-third of all Bitcoin – about 6.26 million BTC – resides in addresses that could be compromised by advanced AI or quantum computing techniques. Of that, 4.33 million BTC are considered exposed due to address reuse, meaning a simple switch to fresh wallets would remove the immediate risk, according to a Glassnode co-founder.

Ethereum researcher Justin Drake has also signaled that rapid AI progress may break the elliptic curve digital signature algorithm (ECDSA) sooner than previously thought, urging users to migrate to new addresses. However, some industry voices, such as Dragonfly’s managing partner, argue that constantly moving tokens does not address the fundamental vulnerability and label the concern as “cryptographic doomerism.”

Industry reactions

The alert has sparked a debate among security specialists and blockchain developers. Some see the scenario as a catalyst for accelerating migration to post-quantum-resistant cryptography, while others caution that the threat remains speculative and that existing best practices—like avoiding address reuse—already mitigate much of the exposure.

Why it matters

If AI-driven worms achieve the ability to infiltrate IoT hardware and target cryptographic keys, the decentralized finance ecosystem could face a new class of attacks that bypass traditional network defenses. The potential for widespread, uncontrolled propagation underscores the need for stronger device security standards and a proactive shift toward cryptographic algorithms that can withstand future AI or quantum capabilities.