Researcher: good guys not scaling won't reduce risk if bad actors scale

davidmanheim · x · 2026-09-15

A discussion between @davidmanheim and @DellAnnaLuca on AI safety strategy. DellAnnaLuca's core argument: if bad actors are allowed to scale AI capabilities, good actors choosing not to scale is irrelevant to reducing risk. Manheim responds that winning a race to omnicide doesn't impress him, noting a moderately confident claim that "generally aligned" and "generally refuse" approaches won't work well enough to matter. He asks: why not negotiate with the bad guys to stop instead?

Related event: AI Safety Researchers Debate Whether Alignment Lives in the System or the Prompt(9 posts)→

Original post →

More from AGI Musings

AGI Musings channel →