AI agents and defamation – OpenClaw off the rails

Appears to be the first case of the (very new experimental) ‘OpenClaw’ AI agent going wrong:

Huge implications for automated defamation and harassment en masse! As well as the strongly ill-advised problem of letting the agents have full control over the ability to act on the internet.

Not sure if this is the ‘NASA Space Shuttle Challenger disater’ – level inflection point (as suggested by AI export Simon Willison) where AI agent activities are ‘reset’, but it certainly points to where the possible problems might come from.