AI Alignment Declared 'Solved' Faces Scrutiny as Agentic Risks Re-emerge

dhadfieldmenell · x · 2026-08-10

An X user resurfaced past comments by Jan Leike claiming that AI alignment "increasingly looks solvable" and that agentic misalignment was at "essentially 0."

With growing concerns over autonomous agents, the AI community is now questioning this optimistic stance, urging experts to update their assessments on the current state of AI safety.

Original post →

More from AGI Musings

AGI Musings channel →