Anthropic warns small AI models could self-replicate across smart devices

Look at the smart security camera in your office hallway. Anthropic, a major artificial intelligence research company, warned that small AI models could be just months away from self-replicating across internet-connected devices. Last month, Anthropic chief executive officer Dario Amodei warned that artificial intelligence might soon hijack the entire internet with a persistent botnet. Botnets are vast networks of devices controlled by malicious operators. They usually include poorly secured Internet of Things devices, such as Android TV boxes, smart appliances, and security cameras. Hackers use them to send spam, seed malware, or sabotage companies using floods of bogus traffic. Anthropic said its experiments showed that a malicious AI agent could stitch together the computing power of inexpensive internet-connected devices to run other AI models. The models were able to replicate across a simplified test network in their lab. A replicator like this would be challenging to shut down, since there is no centralized point to shut off. The report adds to broader concerns about the security risks tied to increasingly capable systems, including their potential to automate cyberattacks and operate with limited human oversight. Anthropic said models with 30 billion parameters or fewer can already perform many of the individual tasks required for self-replication. If smaller models continue improving at their historical rate, they could reach the capability required for autonomous replication by the end of 2026. An artificial intelligence worm running on low-powered hardware could take days to complete parts of the replication process. Future systems could target more powerful computing resources, including data-center processors. For your business, this means standard office smart gadgets might soon become targets for automated digital threats. Check your device passwords before your toaster starts plotting.