Parv Mahajan(@parvmahajan0):The chem/bio automated evals for Mythos 5 are probably underestimating capabilities because they're only giving the models 1M tokens. This doesn't inspire confidence in the risk determinations, and makes me pretty nervous about a lack of 3rd party auditors for AI biorisk.

Parv Mahajan

@parvmahajan0

@METR_Evals | @mcnairai | 🇺🇸

Joined August 2025

165 Following 793 Followers

Parv Mahajan@parvmahajan0

2026.06.24 15:35

The chem/bio automated evals for Mythos 5 are probably underestimating capabilities because they're only giving the models 1M tokens. This doesn't inspire confidence in the risk determinations, and makes me pretty nervous about a lack of 3rd party auditors for AI biorisk.