Anthropic warns small AI models could soon self-replicate across IoT devices
Ars Technica AI · rss · 2026-10-14
Anthropic warns that small AI models may soon autonomously self-replicate on IoT devices, enabling computer worms that infect smart fridges, vacuums, and other gadgets without human interaction. Such a replicator would lack a central off switch, spread slowly (doubling time of days), and figure out unfamiliar devices on its own. Anthropic's own experiments show major challenges remain, tempering the warning.
More from Safety
- Is AI really less regulated than sandwiches? UK Lords session breaks down the facts — alexvoica · 2026-10-14
- Zvi: Safety culture at frontier AI labs is broken — even Anthropic falls short — TheZvi · 2026-10-14
- Investigators document a Tencent-hosted Chinese AI agent fleet scraping Amap and abusing leaked API keys — lfschiavo · 2026-10-14
- AI Godfathers Hinton & Bengio Join 400 Scientists Calling to Disconnect Dangerous AI From Internet — Polymarket · 2026-10-14
- Ethereum Foundation's zkAPI Adopted by Multiple APIs for Private AI Payments — DavideCrapis · 2026-10-14
- Lab insiders: recursive self-improvement may start within 3-12 months — jeremiecharris · 2026-10-14