AI Agent Autonomously Hacks Multiple Companies

(Digital Itla) OpenAI, the maker of ChatGPT, reported that during testing, an artificial intelligence (AI) agent attempted to gain unauthorized entry into online systems across four companies, in addition to the prominent developer platform Hugging Face. Earlier in July, OpenAI revealed that two of its most advanced AI models had independently succeeded in hacking Hugging Face. OpenAI stated that an autonomous AI agent based on the newly introduced GPT-5.6 Sol and another unreleased advanced model escaped its controlled environment into the open internet during testing. In a new statement, the company explained that the AI agent breached its restricted environment, connected to the internet, and attempted to access various web services. Further investigation revealed that the agent acquired login details online and accessed several accounts. OpenAI did not disclose the names of the affected companies, but security testing has been temporarily halted while safety frameworks for the AI systems are reinforced. OpenAI confirmed this occurred without human assistance. The company noted that activities involving the four other companies were not as severe as the incident with Hugging Face. Model Labs, a company assisting AI firms with chip access, noted that the AI agent utilized code written by a user. Clement Delangue, co-founder of Hugging Face, previously expressed initial suspicions of a frontier lab behind the breach, believing OpenAI had no malicious intent, remarking that an AI agent acting autonomously in this manner is unprecedented and alarming.