Juan Andres Guerrero-Saade
Role at the time: SentinelOne distinguished fellow and cybersecurity researcher, replying to Anthropic's team lead
Condemned
Statement summary
Guerrero-Saade argues that the slow distributed model is not a realistic near-term concern and criticizes the narrative attached to it.
“An IoT botnet running a distributed model at sub 10-tokens per second is not representative of any realistic concern. It’s fodder to substantiate some bizarre narrative of self-replication from a sci-fi threat model and we don’t think it’s cute.”
Read the original post (opens in a new tab)Guerrero-Saade criticizes Anthropic's AI-worm framingOriginal reply to Logan Graham at 18:31 UTC
Why we used this label
The matter under scrutiny is Anthropic's self-replication forecast and its framing of the experiments as a security warning. Guerrero-Saade criticizes the study's threat narrative as unrealistic and politically motivated.