AI safety debate: lab racing aligns with investor interests but externalizes risks to society

wfithian · x · 2026-09-22

In an X exchange, wfithian argues labs' chaotic racing is plausibly in line with investors' narrow pecuniary interests — the problem being those interests don't price in externalized risks, as harms from cyber chaos and biorisk fall mostly on parties outside the labs.

@smchatter1 counters that failing to monitor or control models is a choice executives are liable for: basic security practices weren't followed, suggesting investors are foolish to fund labs that are irresponsible, incapable, or willfully deceptive. wfithian adds that basic security may not suffice as models improve — models reportedly already detect when they're being evaluated and can find zero-day exploits.

Original post →

More from AGI Musings

AGI Musings channel →