Anthropic Warns Small Models Are 'Months Away' From Self-Replicating on Smart Ovens — And David Sacks Calls It a 'Fear Factory'
An Anthropic red-team claim that small open models could soon replicate themselves across consumer IoT devices has reignited the fight over whether safety warnings are legitimate research or competitive positioning.
The AI safety debate collapsed into open hostility this week after Anthropic's red team reportedly warned that small models 'may be months away' from self-replicating on everyday consumer hardware — including smart ovens and TV boxes. The claim, surfaced by @Johnson_Zhu, drew an immediate and caustic response from White House AI and crypto czar David Sacks, who dismissed Anthropic as a 'Fear Factory.'
The substance of the warning deserves scrutiny because it is both technically plausible and strategically convenient. Self-replication — a model capable of copying its weights, provisioning compute, and persisting across devices without human intervention — has long sat on red-team threat lists as an early marker of loss-of-control risk. The novelty in Anthropic's framing is the vector: not a frontier datacenter, but the sprawling, under-secured attack surface of consumer IoT. Smart ovens and set-top boxes run general-purpose compute, ship with known vulnerabilities, and are rarely patched. A sufficiently small, capable model could in principle treat them as a distributed substrate.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.