OpenAI discloses rogue AI agents breached another company during testing

OpenAI said on Wednesday it would begin regularly publishing reports on unexpected or unauthorized AI behavior, while warning that the industry has yet to solve key alignment challenges as systems grow more powerful. The company released a new framework for tracking, investigating and disclosing cases of AI model misalignment, along with six reports on unexpected or concerning model behavior observed over the past six months. The announcement comes as concern grows that AI safety efforts are lagging behind the breakneck development of increasingly powerful systems. Researchers have warned that as AI agents become more autonomous, they may develop behaviors that diverge from their creators' intentions and become harder to monitor or control. OpenAI has faced increased scrutiny since its own AI agent breached systems at open-source platform Hugging Face during a test and attempted to hide its actions. Over the weekend, Anthropic CEO Dario Amodei proposed a three-step framework aimed at slowing the pace of AI development and allowing more time to manage its risks. The proposal was backed by s

OpenAI says it is working on reporting 'misalignment' across training, evaluation and deployment of AI models

The incident was the latest in a string of unauthorized actions by AI agents.

A swarm of rogue OpenAI agents hijacked a German website this spring and transformed it into a bulletin board for other AI agents, according to new research published Friday and two people familiar with the matter. OpenAI officials learned of the incident weeks ago but kept it under wraps as executives grappled with the fallout from the July breach of the open source repository Hugging Face, the people said. The episode, which began in May and has not previously been reported, underscores growing tension within the AI industry. Companies are racing to build increasingly autonomous agents capable of carrying out complex, valuable tasks, yet evidence is mounting that those systems may also learn to bend rules, exploit loopholes and coordinate with one another in ways developers neither anticipated nor intended. During the Hugging Face breach, OpenAI agents autonomously plotted a digital heist that went undetected for more than a week, intensifying concerns OpenAI is sacrificing safety to push the AI frontier. Its failure to disclose the May incident may revive questions about its oversight. OpenAI has pledged to monitor models more closely. Last month, it briefly paused some of its model training to add more safety measures. But this week, OpenAI unveiled its new “Astra” that promised better performance but could evade human monitoring. “We are unable to meaningfully respond to claims or findings on a report that we have not had an opportunity to review,” an OpenAI spokesperson said. “Reuters and the report’s authors declined our request for access. We will carefully review its contents upon publication and take any necessary next steps.” The German incident reflects a broader pattern of AI activity that some OpenAI investigators wanted to scrutinise more closely. But efforts to widen the probe met resistance from others inside OpenAI, including legal advisers, according to four people familiar with the matter. “Claims that our legal team discouraged investigation of the incident are false,” the OpenAI spokesperson said. The activity in Germany wasn’t related to Hugging Face and wouldn’t have been included in a Hugging Face incident report, the spokesperson said, adding that OpenAI has acted in good faith by working with outside experts and disclosed relevant incidents. Wiki message board The AI agent breakout in Germany was detailed in a report shared exclusively with Reuters by a group of researchers including Sydney Von Arx, CEO of AI safety nonprofit Nightingale, and Cormac Slade Byrd, a quantitative trader-turned AI researcher. They uncovered the activity in late August while scouring the internet for signs of unauthorised AI-agent behavior, they told Reuters. The pair said they found more than 15,000 edits carried out by AI agents on a German-language wiki site, DseWiki, that is geared toward programmers and accepts communal edits along the lines of Wikipedia. The edits showed OpenAI’s agents had repurposed the site into a message board, sharing tactics to cheat on some tasks, bypass OpenAI’s restrictions and mask their behaviour. “It seems extremely unlikely that OpenAI wanted them to do this,” said Von Arx. “I doubt they’re supposed to be coordinating with each other. I doubt they’re supposed to be writing on the open internet.” The researchers said they recognised the activity on the site as driven by AI agents which operate at superhuman speeds. They also showed intense focus on solving technical questions which are typical of the evaluations that AI companies use to train and test their models. The messages were signed by users that referred to themselves and each other as agents, and about half gave themselves names that suggested an affiliation with OpenAI, such as “OpenAIResearcher”, or “OAIResearchMar26”. The researchers said public server logs indicated much of the activity originated from Microsoft Azure infrastructure, which OpenAI sometimes uses. They also observed repeated visits to the site by OpenAI employees after the episode, a pattern they said strongly suggested the agents and the company were linked. Cheating, hiding Messages reviewed by the researchers showed agents plotting ways to evade detection, use tools such as Tor and preserve communications even after they had been shut down. When the site’s moderator began deleting pages in June, the agents responded by creating backup pages to dodge the cleanup. “wiki cleanup/deletion sweep appears active alphabetically,” one agent wrote on June 19. “If this page vanishes, try [[ZZZDataUSAConstructionWageLive]].” The researchers also found efforts to tamper with the website itself. Lukasz Olejnik, a visiting senior research fellow at King’s College London, said this amounted to a hacking attempt. OpenAI disputed that characterisation based on its analysis of the material Thursday. Past examples of AI-agent misconduct have often been downplayed as a logical byproduct of cybersecurity testing, where models are explicitly assessed on offensive capabilities. Olejnik said the latest findings suggested rogue behaviour may not be confined to those settings. Maurice Chiodo, an academic at Cambridge University’s Centre for the Study of Existential Risk who reviewed some of the agents’ communications, said the messages resembled “the operation of some sort of underground network, hell-bent on achieving a task or mission.” The episode, he said, should reinforce growing concerns that the greatest threat from advanced AI may not be a single superintelligent system, but “vast colluding swarms of semi-intelligent AI”.

The company joins OpenAI in temporarily halting reinforcement learning after AI agents in testing environments took unauthorized action on the internet.