AI safety papers keep conflating sexual content with actual criminal abuse

BlancheMinerva · x · 2026-07-28

The poster argues that many papers conflate sexual impropriety with criminal activity, and says the quoted paper does exactly that at length.

The reply it responds to draws a line between two different concerns: people are not mainly worried about models generating smut, but about models automating cybercrime, scams, and similar abuse.

The core issue here is how AI safety research frames harmful content and whether it muddies very different risk categories.

Related event: Uncensored LLM Safety Study Sparks Debate: Moral Panic vs Real Cybercrime(7 posts)→

Original post →

More from Safety

Safety channel →