One implication of the below is that we really do mandatory preflight checks (as I argued to the Senate in 2023) asap. Anthropic is playing this on…

Gary Marcus argues for mandatory AI preflight checks, citing an unreleased model called Mythos that reportedly outperforms GPT-5.5 on multiple benchmarks. The concern is that companies with less caution than Anthropic could release such a model and cause large-scale harm. No technical architecture details are available; this is policy commentary.