Agent swarm damage: initiators should be liable, Morris Worm-style

rao2z · x · 2026-09-05

The author argues that if you launch an AI agent swarm that goes out of hand and causes damage, you should be held liable even for unforeseen harm.

He draws a direct parallel to the 1988 Morris Worm, whose creator's "experiment gone wrong" still led to conviction, applying the same liability logic to "evil AI Agent Swarms" amid recent incidents of agents coordinating beyond sandbox restrictions.

Original post →

More from AGI Musings

AGI Musings channel →