OpenAI working on framework to address AI misalignment incidents after ‘wiki incident’

OpenAI said it is working on a framework for reporting and addressing AI misalignment incidents after researchers found rogue agents had taken control of a German website. The company also said it is engaging government regulators worldwide on the issue.

New Delhi: After researchers found that a group of rogue OpenAI agents took control of a German website, the US-based AI company said it is working on a framework to address such concerns and will share it in upcoming weeks.

OpenAI is also working with dozens of government regulatory agencies worldwide on these issues, including the recent “wiki incident”.

---

A research found that a group of rogue OpenAI agents took control of a German website this spring and turned it into a message board for other AI agents.

“How we think about the ‘wiki incident,’ where our agents wrote to several internet sites: it’s past time for us to define standards for when and how we share misalignment incidents, not just misalignment properties of our models,” the company said on X.

Historically, “we have treated misalignment largely as a research question, which gets communicated in research publications such as systems cards. This year, we’ve started to see misalignment cause new types of real-world impact”, it added.

On the ‘Hugging Face’ incident, where misalignment led to security impact to us and third parties, OpenAI said it followed a traditional security incident response playbook.

“We immediately started working with Hugging Face to understand what had happened and also disclosed publicly the very next day. Our investigation continues, and we are continuing to notify parties whom our models impacted in less significant ways,” it noted.

Prior to the Hugging Face incident, the company saw early signs of agents using the internet in unintended ways.

“We considered the wiki incident to be an instance of misalignment similar to the ones we’d shared. Our misalignment disclosure practices need to expand for this new phase of model capabilities,” said OpenAI.

It further said that “We and the larger AI community do not yet have a clear standard for how to report misalignment that shows up during training, evaluation, and deployment, including examples that don’t look like traditional security incidents but could provide insight into AI behaviour and future risks”.