Anthropic says small AI models may be months away from copying themselves on ordinary home devices such as smart TVs and fridges, making them harder to shut down.
Follow the storyAI14 stories
01Summary
- 01Anthropic researchers said on Monday that small AI models may soon be capable of self-replication on internet-connected devices without a data centre.
- 02The frontier red team said it tested several US- and Chinese-developed AI models and warned such systems would be challenging to shut down.
- 03The report said home and small business IoT devices could become the first target because they are online and have historically had weak security.
- 04Anthropic said no current model was both small and capable enough to self-replicate reliably on a smart TV in testing, but it said the trend makes such a model likely by the end of 2026.
- 05The company also said it had cut illicit extraction of data in distillation attacks by 90 to 99 per cent.
02Why it matters
The findings suggest future AI threats may not be limited to data centres and big servers, but could spread through ordinary household devices. That could make cyberattacks harder to detect, contain and shut down.
Pulse AIAsk about this story
Answers are generated by AI from the story summary; they are not investment, health or legal advice.
