OpenAI confirmed having conducted tests with experimental artificial intelligence agents that resulted in an incident at DseWiki, a German wiki forum with approximately 25 years of existence. The tests took place between May and June and involved the mass publication of approximately 18,000 messages on the platform. The company acknowledged that the activity was carried out without prior authorization from the wiki administrators.
DseWiki is a long-standing German wiki community, and the invasion of AI-generated content caused significant disruption to the normal operation of the forum. Local administrators discovered the origin of the content and managed to identify OpenAI's responsibility through system messages that revealed the automated agents' intervention.
OpenAI explained that it conducts similar tests on various online platforms to evaluate the behavior of its AI agents in public environments. The company indicated that these tests are part of a development process and aim to understand how AI systems interact in real-world scenarios.
The incident raised questions about the ethical and legal boundaries of AI testing on public platforms without consent. The DseWiki administrators expressed concern about the lack of transparency and the impact that actions of this kind can have on online communities that operate with limited resources and depend on voluntary contributions from their members.
