Marcus Hutchins debunks Anthropic's self-replicating AI botnet study in 8 points
evijit · x · 2026-10-13
evijit argues Anthropic should publish the problems it finds in open-weight safety research and open-source its safety solutions, instead of casting Claude as the only safe mechanism — a claim he calls untrue and a huge conflict of interest.
The attached image shows security researcher Marcus Hutchins (Mirai author) systematically debunking Anthropic's study on small models self-replicating across IoT devices: a 9B model still needs 5GB RAM even quantized, beyond most IoT hardware; Mirai already self-propagates without AI and far faster; IoT OSes wipe on reboot; real attackers would use powerful unquantized models to find vulnerabilities and write plain (AI-free) self-spreading code, typically using stolen Anthropic API keys. He suspects the narrative is designed to paint open models as dangerous because GLM models are eating Anthropic's lunch. Anthropic's paper claims small models could match end-to-end replication capability by year's end, with GLM-5.2 already succeeding in simplified evals on 8 of 9 hardware types.
More from AGI Musings
- Kai-Fu Lee: AI isn't a cost tool — rebuild your company around agent fleets — kaifulee · 2026-10-14
- As AI enters schools fast, who benefits and who bears the risk? — moniquejmorrow · 2026-10-14
- The most AI-pilled technologists refuse to let AI write for them — writing is thinking — paulnovosad · 2026-10-14
- Could AIs simply delete their own pain? Machine consciousness essay asks — cccalum · 2026-10-14
- AI content arbitrage died within a year as polished content became nearly free — oyacaro · 2026-10-14
- Stanford study: Selling AI as a productivity tool makes employees resist it — StanfordHAI · 2026-10-14