Ex-Microsoft engineer: AI safety needs no doom talk — bad security is just bad business

gordic_aleksa · x · 2026-09-30

Aleksa Gordic argues labs don't need doomer narratives or global government to pace the frontier: if an AI lab can't get supporting software like sandboxes right, they risk leaking IP (including model weights), reputational damage and lost trust — it's simply bad business, and free markets go far. He adds the real task is making models themselves safe, aligned and monitorable.

Related event: Ex-Microsoft Engineer: Market Incentives Alone Can Keep AI Labs Safe(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →