Anthropic has warned that artificial intelligence could make it easier to create malware capable of spreading across smart devices. The concern is not that AI has suddenly made every connected camera or thermostat vulnerable. It is that tools that help people write and adapt software could lower the effort needed to build attacks that target the often-overlooked devices on home and business networks.

Smart devices are attractive targets for a familiar reason: many stay connected for years, while their software receives less attention than the software on a laptop or phone. A router, camera, printer, or smart speaker may be running outdated firmware, using a weak or unchanged password, or exposing an administrative service to a network where it does not need to be.

What AI could change

Malware authors already automate parts of their work, from scanning for exposed devices to trying known vulnerabilities. AI could help speed up tasks such as adapting code to different environments, making sense of technical documentation, or producing scripts for repetitive steps. That may let less experienced attackers attempt work that previously demanded more specialist knowledge.

But “AI-assisted” should not be mistaken for “self-running” or “unstoppable.” A model does not automatically give malware access to a device, discover an unknown flaw, or make incompatible hardware behave alike. An attack still depends on conditions such as a reachable device, a usable weakness, stolen credentials, or a user being tricked into taking action. Smart-device platforms also vary widely, which makes reliable, broad infection harder than the headline version of the risk might suggest.

The realistic concern is that automation could help attackers test more opportunities and tailor familiar techniques to more devices. If compromised devices share a network, an attacker may also try to move from one device to another. How far that works depends on the network’s design, device protections, and whether each target has a weakness the malware can exploit.

Why connected devices can be hard to defend

People do not always know how many devices are on their network. Some are tucked away behind televisions or installed by a service provider, and their update settings are easy to overlook. Others stop receiving security updates while remaining plugged in and online. A device can therefore become a weak point even when its owner keeps their computers up to date.

That does not mean every smart device is an imminent threat. It does mean that basic maintenance matters. Check which devices are connected to your router, install available firmware updates, and change default passwords. If a device no longer receives updates and handles sensitive activity, replacing it may be safer than leaving it exposed indefinitely.

• Use unique passwords for device accounts and the router’s administrator login.
• Turn off remote administration and other internet-facing features you do not use.
• Separate smart devices from computers and phones where your router supports a guest or IoT network.
• Remove unused devices from the network rather than leaving them connected without a clear purpose.

For organizations, the same issue calls for a current device inventory, a plan for unsupported hardware, and network rules that limit what connected equipment can reach. Segmentation will not fix vulnerable firmware, but it can reduce the damage if one device is compromised.

Anthropic’s warning is best read as a reason to take the security of connected equipment seriously—not as proof that an AI-driven outbreak is underway. AI may make some parts of malware development and deployment more accessible, while the weaknesses it would exploit remain largely familiar: missed updates, poor credentials, exposed services, and networks that give devices too much access to one another.