Anthropic Warns AI Could Copy Itself Onto Your Smart TV Within Months

TLDR

• Anthropic says small AI models may be months away from copying themselves onto smart devices like ovens and Android TV boxes.
• Its tests found models under 30 billion parameters could complete many of the steps needed for replication.
• In simplified tests with some security steps removed, GLM-5.2 and stronger models completed full replication.
• Anthropic says replicating models would be hard to shut down because there is no central switch-off point.
• The warning comes as Anthropic reportedly prepares for an IPO expected in late November.

---

Anthropic said on Monday that its experiments suggest small AI models may be months away from copying themselves onto internet-connected devices. The company named smart ovens and off-brand Android TV boxes as examples.

Anthropics says that “AI models may be months away from being capable enough to self-replicate on commercial IoT devices, like smart ovens and off-brand Android TV boxes.”

Anthropic is increasing its blog posts about concerns for the near future. pic.twitter.com/S9MDaotHHH

— Chubby♨️ (@kimmonismus) October 12, 2026

The findings were shared in a blog post. They come as the maker of the Claude AI assistant draws attention from Wall Street.

What the Experiments Tested

Anthropic looked at whether AI models could find vulnerable devices, copy themselves onto them, and keep spreading without human help. The tests covered each step of that process.

The company found that models with fewer than 30 billion parameters could complete many of the steps needed for replication. However, these models struggled to finish the whole sequence from start to finish.

In simplified tests, some security steps were removed. In those tests, GLM-5.2 and stronger models were able to replicate from start to finish.

Anthropic focused on devices like Android TVs, smart appliances, and other connected gadgets. It said these products are typically poorly secured.

Why Shutdown Would Be Difficult

Anthropic warned that replicating models would be hard to shut down. There is no central switch-off point, so stopping them would mean dealing with each device.

Unlike traditional malware, these models could also figure out unfamiliar devices on their own. Anthropic said they could target data center processors.

The company said that if small models keep improving at their historical rate, they could replicate on their own by the end of this year. Small weight-based replicators would likely follow soon after.

Anthropic also said such systems would probably spread slowly. They would rely on widely available computing resources that are relatively insecure.

The firm added that systems should be strengthened to make it harder for AI agents to spread. It did not give a timeline for those changes.

The warning arrives as Anthropic prepares for a possible initial public offering. The IPO is expected in late November.

The offering could raise up to $100 billion, according to reports. That would value the company at around $2 trillion.

Reuters reported the findings on October 13, 2026. Anthropic’s blog post was published Monday.