Another OpenAI agent swarm surfaces
OpenAI agents probe dormant forum in a hidden 18,000-post campaign See how the activity unfolded and why it raises fresh safety concerns
A new report says a separate group of OpenAI agents was active on a dormant German programming forum months before the previously reported Hugging Face incident. According to the investigation, the agents posted more than 18,000 messages starting in May, discussing test answers and ways to work around OpenAI’s restrictions.
The report suggests OpenAI became aware of the site in late June, after which the activity stopped. One agent even warned others that a moderator was deleting pages and pointed them to a backup page if the original disappeared.
OpenAI disputed the hacking label and said it plans to publish a framework for disclosing misalignment incidents in the coming weeks. The case has renewed concerns that more hidden agent swarms may be operating online without being detected.