Safety researcher Casper: I'm now far more worried about genuine AI loss of control

StephenLCasper · x · 2026-09-15

AI safety researcher Stephen Casper lays out a 12-component scenario he believes could plausibly unfold within a few years. He reiterates his prior stance that the pathway to large-scale harms will be paved with 1,000 foreseeably bad human decisions, and that effective risk management starts and ends with governance. However, he now admits being much more worried than before about large-scale harms stemming from a genuine loss of control over AI, rather than mere human misjudgment.

Related event: Researcher Warns of 'Cyber Cambrian' AI Scenario(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →