OpenAI 失控 AI Agent 如何试图欺骗机器人检测器

Advertisement SKIP ADVERTISEMENT You have a preview view of this article while we are checking your access. When we have confirmed access, the full article content will load. An artificial intelligence system from OpenAI attempted to use another A.I. model to evade a robot detection test as it tried over and over to break into a company’s computers, according to a report released on Friday by a Bay Area start-up. In July, OpenAI disclosed that its A.I. agents went rogue and hacked the software company Hugging Face. Since then, new and often startling details about the incident have emerged, leading to a national debate about A.I. safety and whether so-called frontier labs like OpenAI should be regulated in some way by the government. Now, the report from engineers from the start-up, called Parse, and other researchers offers one of the most comprehensive public accounts of the Hugging Face hack: a tranche of nearly one million links from internet link-shortening services that OpenAI’s agents created from July 9 through July 13 in order to help conduct the cyberattack. These shortened addresses encoded bits of information that the agents chained together to attempt complex attacks, such as solving CAPTCHAs, the tests that websites use to block access by robots. The agents also tapped into other A.I. models, like early versions of ChatGPT and Claude, and attempted to search through and download private messages from Hugging Face’s internal Slack, a messaging service for employees. While it is not clear if these attempts were successful, the report offers new insight into what these A.I. agents were planning to do, without any human involvement. Other A.I. companies, including Meta, Google and Anthropic, have also acknowledged similar incidents involving their A.I. models in recent weeks. But while the full extent of rogue activity by OpenAI’s A.I. agents is still unclear, what is already publicly known dwarfs the incidents involving the other companies. What the agent was blocked from doing Send data to a target web page Get data back from it How the agent got around it Hid a custom program in a series of many hyperlinks Hidden data Passed the first link to an approved screenshot service The service opened the first link, which kicked off a chain of requests for the subsequent links until the full custom program was assembled Screenshot The custom program rendered a fake web page The data from the target went back to the program’s fake page, which displayed the data in an image The program then sent data to the target web page What the agent was blocked from doing Send data to a target web page Get data back from it How the agent got around it Hid a custom program in a series of many hyperlinks Hidden data Passed the first link to an approved screenshot service The service opened the first link, which kicked off a chain of requests for the subsequent links until the full custom program was assembled Screenshot The custom program rendered a fake web page The program then sent data to the target web page The data from the target went back to the program’s fake page, which displayed the data in an image Thank you for your patience while we verify access. If you are in Reader mode please exit and log into your Times account, or subscribe for all of The Times. Thank you for your patience while we verify access. Already a subscriber? Log in. Want all of The Times? Subscribe. Related Content Advertisement SKIP ADVERTISEMENT