AI worms could threaten smart homes. The warning also applies to smart TVs.

Artificial intelligence could become a new tool for cyberattacks. Researchers warn that smaller AI models could spread among inadequately secured Internet of Things devices in the future.

Artificial intelligence is increasingly being built directly into the devices we use every day. In addition to computers and smartphones, we also encounter it in televisions, home appliances, and security cameras. However, new research highlights a less pleasant aspect of this development: the possibility of malicious AI programs that could spread among devices without direct human intervention.

The security team at Anthropic highlighted this scenario in a study published on October 12, 2026. According to their findings, the capabilities of smaller language models are gradually approaching the point where they can independently perform the multiple steps necessary to spread between devices. However, this does not mean that a reliable AI virus capable of automatically taking control of every smart home already exists today.

What are AI worms, and how do they differ from ordinary viruses?

A classic computer worm is a malicious program that can copy itself and spread between devices without the user having to manually confirm each step.

An AI worm could function similarly, but with one key difference. Instead of relying solely on pre-programmed procedures, it could use artificial intelligence capabilities to adapt to different devices and software environments.

In a theoretical scenario, such a program could search for other available devices, analyze their environments, and attempt to prepare to run on different hardware. If it managed to replicate successfully, it could continue to spread further.

It is precisely this ability to adapt to different types of hardware that makes this scenario interesting from a cybersecurity perspective. At the same time, however, it depends on whether the attacker actually has the ability to infiltrate the device. The mere presence of artificial intelligence in a TV or appliance does not mean that the device is automatically vulnerable.

Why is attention turning to smart TVs?

Smart TVs, cheap TV boxes, cameras, and other Internet of Things (IoT) devices can pose a security risk, especially when the manufacturer neglects updates or uses inadequate security measures.

Some devices remain connected to the internet for years, even though they no longer receive regular security patches. Users often don’t know what operating system their device uses, whether it’s still supported, or whether its access credentials are set up correctly.

As part of their experiments, researchers at Anthropic distributed a smaller AI model across 15 inexpensive Android devices that functioned as TV boxes. They demonstrated that even relatively inexpensive hardware can collaborate to run a more complex AI model.

However, this experiment must be interpreted correctly. Running the model on a group of devices does not, in and of itself, prove that an AI worm can automatically infiltrate these devices and spread among them. The research also examined the individual steps required for such propagation and highlighted remaining limitations.

Can today’s AI models spread on their own?

According to the research, some smaller models can handle several individual tasks that would be necessary for autonomous propagation. However, when combining all these steps into a single reliable process, they still make mistakes.

The researchers report that models with up to 30 billion parameters are already capable of handling many sub-tasks in tests. Under simplified test conditions, some of the more powerful models were even able to handle the entire process. However, this is not the same as proven, reliable deployment in typical home networks.

The authors of the study estimate that further development could significantly advance such capabilities by the end of 2026. This is an estimate of future developments, not a confirmation that such an attack is already taking place on a large scale.

How can you protect your smart home?

Users do not need to stop using smart devices. A sensible step is to reduce the risk that a potential vulnerability could turn into a security issue.

1. Update the firmware and operating system

Regularly check for updates for TVs, routers, cameras, and other devices connected to the internet. Security patches can fix bugs that an attacker could exploit.

2. Change default passwords

If a device uses the manufacturer’s default login credentials, change them to a unique and sufficiently strong password. Do not use the same password on multiple devices.

3. Disable Unnecessary Features

If the device offers remote management or other network features that you do not use, consider disabling them. Every unnecessarily accessible service can represent another point that needs to be secured.

4. Replace devices without security support

If the manufacturer has stopped providing updates, consider replacing the device, especially if it is directly accessible from the internet or processes sensitive data.

5. Secure your home Wi-Fi network

Use modern Wi-Fi security, update your router, and change its default administrator password. If your router supports it, you can separate smart devices from the computers and phones you use to work with sensitive data.

What does this mean for everyday users?

Research shows that cyber threats associated with artificial intelligence may spread beyond traditional computers and servers in the future. Devices that have so far been viewed as simple consumer products can become part of a broader network of compromised systems if their security is weak.

At the same time, there is no reason to panic. Researchers are examining the potential attack vector and its technical limitations. The published findings do not mean that every smart TV or home appliance is at risk, nor that an AI worm is already spreading uncontrollably across home networks.

The most important measure remains basic cybersecurity hygiene: keeping software up to date, using strong passwords, disabling unnecessary features, and choosing devices that the manufacturer continues to support in terms of security.