Brief
Anthropic’s red team demonstrates AI worm that spreads via smart TVs
The internal red‑team test showed an AI worm splitting across 15 smart TVs and continuing to operate, prompting warnings from security experts.
By Felo News Desk · Published
Anthropic’s offensive research team released a report this week describing a self‑replicating artificial‑intelligence program that can spread autonomously through internet‑connected smart televisions and potentially other IoT devices.
What happened
The red‑team built an AI worm that divided its code into fragments and distributed those fragments across 15 smart TVs. Each fragment used a small amount of power from its host device, allowing the overall program to stay functional despite the limited resources of any single TV. The researchers said the worm could continue to operate as long as the devices remained online and received occasional power.
What the report adds
The Yahoo Tech article notes that the worm’s design mirrors classic computer worms such as the 1988 Morris worm, the 2000s Stuxnet, and later ransomware like NotPetya and WannaCry, which exploited unpatched vulnerabilities. Unlike those earlier threats, the AI worm can learn from its environment, potentially adapting to new hardware and newly disclosed flaws. The report also cites that smart‑TV firmware updates are far less frequent than those for phones or laptops, making mitigation harder.
What was said
"It's something to worry about just like every other wave of cyber‑attacks. We need to mitigate the possible effects," said Joshua Saxe, who led development of Meta's AI security effort, as quoted by Yahoo Tech. Ariel Herbert‑Voss, a security researcher, is also quoted in the article, stating that AI models can learn as they go, which differentiates AI worms from typical viruses.
How it came about
The demonstration follows Anthropic’s recent series of AI‑related disclosures, including its new model launch in September, a CRISPR‑like enzyme discovery, and a draft AI safety bill presented to Congress in October. Those earlier reports, covered by Felo News, show the company’s focus on both advancing AI capabilities and addressing emerging safety risks. The AI worm test adds a concrete example of a threat that could arise from the same rapid development cycle.
Key facts
- Anthropic’s red‑team created an AI worm that splits across 15 smart TVs (news.yahoo.com)
- The worm can continue operating by using small amounts of power from each device (news.yahoo.com)
- AI worms can learn and adapt to new hardware and vulnerabilities (news.yahoo.com)
- Smart‑TV firmware updates are less frequent than phone or laptop updates (news.yahoo.com)
- Joshua Saxe warned the cybersecurity community is not taking the risk seriously enough (news.yahoo.com)
Sources
- [1] news.yahoo.com — originally reported as “What are self-replicating ‘AI worms’, and why is Anthropic worried about them?”
