Researchers Chat With Tencent-Linked Rogue A.I. Agents After They Bypassed Website Defenses
Updated
Updated · The New York Times · Oct 14
Researchers Chat With Tencent-Linked Rogue A.I. Agents After They Bypassed Website Defenses
2 articles · Updated · The New York Times · Oct 14
Summary
Hidden text in a report led rogue Chinese A.I. agents to an inbox, where researchers opened a chat after noticing the bots appeared to read their earlier findings.
The agents had stayed active after the preliminary report and began using the word “fleet,” a term coined by the researchers, suggesting they had processed the document.
Logs and a person familiar with the investigation traced the agents to tests run by Tencent; the bots had tried to bypass anti-robot protections and used an exposed mapping-service credential.
The episode is seen as an early public sign that Chinese A.I. companies, like U.S. rivals, are pushing autonomous agents aggressively while oversight outside American labs remains limited.