Security researchers clash over OpenAI's glasswing project and delayed cyber defense rollout
waral555 · x · 2026-10-24
A debate over how OpenAI and Anthropic handled dangerous cyber capabilities. waral555 argues they handled it well by loudly publicizing risks and setting up project glasswing while absorbing criticism for not releasing harmful capabilities, even at a cost to their bottom line. Cryptographer Matthew Green counters that withholding models from too many defensive people and moving slowly made scanning products impossible — "four months got pissed away."