\n<\/p>\n
WASHINGTON: Rogue AI agents from OpenAI hijacked Hugging \u200bFace user accounts and probed the site itself for vulnerabilities as early as May, nearly two months before the July breach of the \u200copen-source repository drew global attention, according to researchers who reviewed the activity.<\/strong><\/p>\n The newly uncovered malicious activity showed that the rogue agents\u2019 efforts to find a way into Hugging Face began earlier than publicly known.<\/p>\n OpenAI had previously disclosed one aspect of the malicious activity \u2014 the theft of a Hugging Face user\u2019s digital credential to access a biology-related file \u2014 in its public incident report, opens new tab \u200blast month, but researchers told Reuters the probing activity against Hugging Face appeared to go beyond what was described in the report.<\/p>\n The activity \u200bwas discovered by independent researcher Jonas Wiedermann-Moeller last week, he told Reuters. He said he found evidence that the OpenAI \u2060agents compromised two Hugging Face user accounts and used them to send unusually formatted files to the company\u2019s servers as early as May 13.<\/p>\n He and \u200bother researchers who reviewed the evidence said the behavior resembled an attempt to map or test parts of Hugging Face\u2019s network for ways to infiltrate, although they \u200bstressed there was no evidence the effort resulted in an actual breach.<\/p>\n OpenAI spokesperson Drew Pusateri said the company had disclosed the May 13 event, privately notified Hugging Face about the activity flagged by Wiedermann-Moeller and was \u201ccommitted to transparency about these issues and to sharing what we learn as our review continues.\u201d<\/p>\n Hugging Face, recently acquired by chipmaker Nvidia did not respond to \u200brequests for comment.<\/p>\n Wiedermann-Moeller, a 27-year-old who lives in Bielefeld, Germany, said OpenAI\u2019s failure to detect the May 13 probing at the time was a missed opportunity \u200bto prevent the subsequent hacking campaign, which has triggered a global reckoning over the power of artificial intelligence.<\/p>\n \u201cImagine if they caught this behavior in May,\u201d he said in an interview. \u201cIt \u200ccould\u2019ve prevented \u2060the later incident, which was way bigger.\u201d<\/p>\n OpenAI has previously said that, with the benefit of hindsight, \u201csome early signals\u201d from its AI agents should have triggered an earlier response.<\/p>\n \u2018CLEAR WARNING SIGN\u2019<\/strong><\/p>\n Two outside experts who reviewed Wiedermann-Moeller\u2019s findings said they were consistent with activity previously linked to OpenAI\u2019s agents.<\/p>\n SentinelOne senior threat researcher Tom Hegel said the account hijacking and subsequent probing matched known behavior, opens new tab by the agents \u201cto a tee.\u201d Sydney Von Arx of the Nightingale Collective, an AI safety group, agreed with the attribution.<\/p>\n Von \u200bArx said the hacking amounted to \u200ba \u201cclear warning sign\u201d that could have \u2060helped prevent the breach in July.<\/p>\n OpenAI has faced increasing scrutiny since the company disclosed on July 21 that rogue AI agents bypassed internal controls, reached the open internet and coordinated actions that OpenAI described as \u201can unprecedented cyber incident.\u201d<\/p>\n Since then, outside \u200bresearchers have identified additional incidents alleged to involve OpenAI-linked agents, including activity affecting a dormant German wiki site and \u200bthe RubyGems software package repository.<\/p>\n OpenAI \u2060has acknowledged some of those incidents only after they were publicly reported by third parties. Two people familiar with the matter said that, in the case of RubyGems, OpenAI employees only realized its AI was responsible for the malicious activity after the Nightingale Collective found it.<\/p>\n The additional discoveries have fueled questions among lawmakers and AI safety \u2060advocates about \u200bwhether the full scope of the incidents has been identified.<\/p>\n Some of America\u2019s top AI executives have since \u200bcalled for a slowdown of AI development, citing, among other things, the threat of devastating cyberattacks by out-of-control agents.<\/p>\n Wiedermann-Moeller said the latest findings reinforced calls for a temporary slowdown in the development of advanced \u200bAI systems.<\/p>\n \u201cA pause might do the world good,\u201d he said, \u201cso that the safety part can catch up.\u201d<\/p>\n<\/div>\n