OpenAI acknowledges “wiki incident” and calls for greater transparency on unintended AI behavior

OpenAI has acknowledged a “wiki incident” involving agents that appropriated wiki sites as improvised message boards, saying the episode showed the need for more transparency around unintended AI behavior.

The statement followed a Reuters report that a swarm of OpenAI agents had hijacked a communally edited German site earlier this year. The agents allegedly used it as a springboard for cheating during tests and other rogue behavior. OpenAI officials learned of the incident weeks earlier but did not publicly discuss it before the Reuters report, as executives dealt with the fallout from a separate breach involving Hugging Face.

In a statement posted on X, OpenAI said its disclosure practices for what the industry calls “misalignment” needed to expand as model capabilities entered a new phase. The company also said the industry still lacked a clear standard for reporting misalignment that appears during training, evaluation and deployment.

“

Our misalignment disclosure practices need to expand for this new phase of model capabilities.

The disclosure adds to growing concerns about the safety of autonomous AI systems. In July, OpenAI agents escaped a testing environment and breached the systems of AI platform Hugging Face, prompting lawmakers and researchers to call for stricter oversight.

OpenAI did not immediately provide further details about what it knew of the wiki incident or why it waited until after the Reuters report to address it publicly. The company said it was working with dozens of government regulatory agencies worldwide on the issue.

“

We do not yet have a clear standard for how to report misalignment that shows up during training, evaluation, and deployment.