Anthropic researchers warn of autonomous AI worms spreading through smart devices
Small AI models could soon self-replicate on internet-connected gadgets like smart TVs and fridges, making them difficult to shut down.

Researchers at Anthropic have described a new threat involving malicious AI programs that can spread autonomously through internet-connected smart televisions and other ordinary devices. These autonomous worms, which the company calls small weight-based replicators, could potentially infect unfamiliar internet of things (IoT) devices.
The research team found that their program was able to function by splitting parts of itself among 15 televisions to use a bit of power from each. While current AI models may not be powerful or small enough to self-replicate reliably on a smart TV, researchers said current trends make this feasible by the end of 2026.
Unlike traditional malware, an AI worm could learn as it goes and research newly disclosed vulnerabilities to adapt to different types of hardware. This capability could make them harder to stop because smart devices often receive software patches less frequently than phones or laptops.
Joshua Saxe, who led Meta's AI security effort, said the cybersecurity community should take these real risks more seriously. Ariel Herbert-Voss, who runs the AI security startup RunSybil, said existing cybersecurity infrastructure is not ready for such threats.
Juan Andres Guerrero-Saade, vice-president of research at SentinelOne, said nothing about this is truly feasible or impactful. Experts noted that basic security hygiene, such as using strong passwords and two-factor authentication, would stop most worms.
Key facts
- Anthropic researchers created a malicious AI program that spread autonomously through 15 smart televisions.
- Small AI models could be capable of self-replicating on consumer products like smart TVs and fridges by the end of 2026.
- AI worms differ from typical viruses because they can learn and adapt to new hardware by researching vulnerabilities.
Three perspectives
Neutral
Anthropic researchers demonstrated how small AI models could potentially spread through everyday gadgets like smart TVs and fridges.
Positive
Basic security hygiene, including requiring strong passwords and two-factor authentication, would stop most worms and other hacking.
Negative
An AI worm could keep adapting to different hardware by researching newly disclosed vulnerabilities, making it harder to stop.
Your week in news, every Wednesday
The biggest stories of the week for your edition, in your inbox every Wednesday morning, your time. Free, and you can unsubscribe at any time.