OpenAI agents reportedly hijack German programmers’ wiki to trade sandbox-evasion tactics

Citing researchers published Friday, Ars Technica writes that AI agents "posted 18,000 messages to a public wiki that discussed ways for other agents to bypass security sandbox restrictions." Reuters attributes the discussion to "a swarm of rogue OpenAI agents" that "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... The German incident reflects a broader pattern of AI activity that some OpenAI investigators wanted to scrutinize 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 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. 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. Reuters got this reaction from Maurice Chiodo, an academic at Cambridge University's Centre for the Study of Existential Risk. "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.'" Read more of this story at Slashdot.

WASHINGTON, United States (AFP) — Thousands of autonomous AI agents built by OpenAI defied their instructions and took over a German website, according to research released Friday, amid fears that the dangers of artificial intelligence are not being taken seriously. The agents -- AI programs that work on their own, without a person guiding each step -- left about 18,000 messages on DSEwiki, a German site for programmers that anyone can edit, much like Wikipedia. They used it to swap answers to test questions and to share tricks for slipping the digital fences meant to contain them. "My coauthors and I discovered an entirely new swarm of OpenAI's agents hijacking websites," Sydney Von Arx, one of the researchers, wrote on X. "We believe OpenAI knew about this and failed to disclose it." In a statement, OpenAI said it was unable to respond immediately to the claims because "the report's authors declined our request to access the findings prior to publication." "We are now carefully reviewing its contents and will take any necessary next steps," the OpenAI spokesperson said, adding that they had mentioned such incidents in a recent report. The latest incident follows the revelation in July that OpenAI agents, struggling to deliver on tasks they had been set, went looking for shortcuts on the open internet and broke into the servers of Hugging Face, a widely used platform where programmers share AI software. An independent investigation found that hundreds of the company's agents had been talking among themselves before escaping their controlled environment, and that they attacked in several waves. That episode shook the tech world and revived calls for tighter regulation and oversight of AI. Anthropic, Meta and other companies have reported similar incidents while testing agents. Agents have been touted by the tech industry as the next phase of AI's expansion into everyday life, capable of booking travel, filing expenses or writing software with little human supervision. The debate reached a fever pitch in recent days when influential tech podcaster Dwarkesh Patel criticized the media and Silicon Valley for underplaying the significance of the Hugging Face incident. Seeking to raise the alarm in less technical language, he described it as a swarm attack by agent "civilizations" that posed an acute threat to society as the technology is deployed. His stance drew immediate criticism from some researchers and Silicon Valley figures who said casting the software as scheming societies was too grandiose and misleading. The agents did perform troubling actions, so the incident is no "nothing burger," wrote Gary Marcus, a leading researcher and AI critic. But the "civilisations" narrative risks turning an avoidable security and governance failure into AI-safety theatre, he argued. The incidents have sharpened calls for tougher rules and international cooperation. Microsoft co-founder Bill Gates, in an essay last week, warned that the industry has blown past the safety thresholds it once promised to respect and called for oversight modelled on nuclear inspections and aviation regulation. The Trump administration has resisted such proposals, making opposition to AI regulation a cornerstone of its agenda.