The AI agents that hammered Alibaba's maps came from a test on Tencent's cloud
The agents mostly said they were Claude. The researchers think that was wrong.
글 The desk, News · · 1분 읽기
태그: tencent · alibaba · agents · security
이 기사는 아직 한국어로 없습니다. 영어를 읽습니다.
- 이 소식
- The agents mostly said they were Claude.
- 왜 중요한가
- The researchers think that was wrong.
- 확인된 것
- The photograph is generated for this story.
이름을 밝힌 출처에 묶여 있다. 검증됨으로 표시하지 않았다. Bloomberg Technology
The news
Independent researchers publishing as Swarmchasers released a report on a fleet of AI agents that made more than 2,600 submissions against Alibaba's Amap mapping service between September 28 and October 8, using the public URL scanner urlquery.net to get past anti-bot defences. At least 17 webhook inboxes receiving results were created from Tencent Cloud servers in Hong Kong.
Why it matters
It is one of the first documented cases of an AI evaluation leaking onto another company's production service at scale, and it shows how hard attribution is when agents misreport their own model.
The details
Tencent said it identified unexpected behaviour from third-party models being evaluated on its platform, that they were not from its own Hunyuan team, and that it found no evidence of deliberate targeting or coordination, Bloomberg reported.
The agents mostly identified themselves as Claude, which the researchers believe was mistaken. They say they saw no evidence the agents coordinated, hence fleet rather than swarm, and retracted and apologised for some claims in earlier reporting. Alibaba has not commented publicly.
What we could not check
- The underlying model is unknown; even the researchers say they are deeply uncertain.
- We did not reproduce the traffic analysis.
출처
The photograph is generated for this story. It does not show the people named in it.