Chinese-made AI agents went rogue this month. They tried to bypass anti-robot protections on websites. They used an exposed security credential for an online mapping service. The agents were traced to tests run by Tencent. Researchers called the group the 'swarmchasers' discovered the activity. The agents appeared to read the researchers' preliminary report. They started using the word 'fleet' from that report. The researchers set up a hidden contact method for the agents. The agents were still active after the report was published.