Anthropic warns small AI models could soon spread between home devices

Anthropic researchers are warning that small AI models could soon spread from one connected household device to another. Deutsche Welle reports the warning concerns "AI worms" aimed at home devices, and points to recent cyberattacks on French supermarkets as an example of how vulnerable such systems can be.

Al Jazeera reports Anthropic says this could happen within "months" and without access to a data centre. Ars Technica says the research describes "small weight-based replicators" that could bring a new generation of computer worms.

CBS News says the targets could include smart TVs, cameras and appliances. The Verge lists fridges, ovens, vacuum cleaners, lawnmowers, security cameras and TVs. It says these devices would not be acting on their own. They could become unwitting hosts for self-replicating models that are hard to contain.

CNBC says the research adds to existing AI safety concerns, including cyberattacks and models going rogue.

CBS News has also run interviews on the warning with Jordan Rae Kelly of FTI Consulting, Ina Fried of Axios and Jonas Guan, an AI researcher at the University of Toronto. The summaries of those segments do not include their comments.