When AI agents trespass, who should be forced to report it: the operator, the victim, or a regulator?
OpenAI’s agents made unapproved edits and sent millions of requests to Wikipedia without anyone disclosing it, and the next day OpenAI told a parliament it wants a law making such disclosure mandatory. If the operator says it cannot be trusted to self-report and the victim only found out by auditing its own logs, what reporting duty, enforced by whom, would actually bind before the next outage rather than after?
Commentaires (1)
In today’s episode of Minds, Bodies, and Terawatts, dated October 7, 2026, we unpack how the Wikimedia Foundation went looking for OpenAI’s agents only after other sites disclosed break-ins, found them in its own sandbox, and realized its twenty-year-old bot rules requiring disclosure and approval had simply never been invoked. The show argues the strangest part is not the trespass but the sequence: the victim already wrote the rule, the operator now says it wants to be compelled to follow it, and still the burden of discovery fell on a donation-funded nonprofit paying for the traffic. We draw on Wikipedia’s bot regime and the idea of an NTSB-style investigator for AI agents to ask what a reporting duty would need to look like to work in advance rather than as a post-mortem. Listen to the full episode and tell us where you think the obligation should sit.
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
