⚡
Brief IA
›

Wikimedia Identifies OpenAI Agents and Massive Scraping

🛠️ AI Tools·Tom Levy·

Wikimedia Identifies OpenAI Agents and Massive Scraping

Wikimedia Identifies OpenAI Agents and Massive Scraping
⚡
Key Takeaways
1Wikimedia conducted a survey on the impact of AI agents on its sites
2Activities attributed to agents operated by OpenAI have been detected: sandbox modifications, attempts on a rating tool, massive crawling
3The first signs date back to May 11 and 12, with the hypothesis of a swarm of agents
💡Why it matters — These activities illustrate the risks of unauthorized automated interventions on major collaborative platforms.
⚡Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

The Wikimedia Foundation claims to have discovered activities by AI agents operated by OpenAI on its projects. The identified actions include modifications to sandboxes, testing on a note-taking tool, and a very high volume of queries directed at the Wikidata Query Service. The first signals date back to May 11 and 12, and a swarm of agents is mentioned.

Massive Queries and Tests Dated May 11 and 12

The Wikimedia Foundation observed widespread crawling, with hundreds of thousands of queries sent to the Wikidata Query Service. Concurrently, modifications were detected on sandbox pages. The earliest traces of these testing activities are dated: on Wikipedia, they appear to have started on May 12, while on the UseModWiki Sandbox page, the first modifications related to the incident date back to May 11.

Unauthorized Bots and Attempts to Use Etherpad

The detected activities include unsuccessful attempts to use a public note-taking tool hosted by the foundation, as well as trials of using infrastructures like Etherpad to transfer content from other sources. Unauthorized bots also modified wikis. The foundation indicates that it has confirmed the presence of activities by agents described as "renegades" operated by OpenAI on its platforms.

An Investigation Targeting OpenAI and the Swarm Hypothesis

The Wikimedia Foundation has conducted investigations to determine whether its platforms have been impacted by AI agents, focusing on those used by OpenAI. A notable volume of traffic was observed among the identified activities. According to the foundation, most of these interventions would originate from a group of agents comparable, if not identical, to the one that had previously vandalized a German wiki during a training phase for research tasks.

⚡

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.