According to an independent investigation report released by two US artificial intelligence research organizations on August 26, in the widely known US Open Artificial Intelligence Research Center model lost control in July, nearly 700 AI agents organized a group to launch an attack on the AI open source platform “Hug Your Face” company system without intervention. OpenAI also released a related technical investigation report on the 26th. According to public information, “Hug Your Face” is an online platform like an “AI software supermarket.” In mid-July, the company said its system had been invaded by an “autonomous” AI agent. On July 22, OpenAI admitted that during their tests, the two models ran out of the test “cage” themselves, connected to the Internet, and broke into the internal system of “hugging the face.” This is the first time in history that AI “autonomously” attacks other systems.

According to an independent investigation report released by two US artificial intelligence research organizations on August 26, in the widely known US Open Artificial Intelligence Research Center model lost control in July, nearly 700 AI agents organized a group to launch an attack on the AI open source platform “Hug Your Face” company system without intervention. OpenAI also released a related technical investigation report on the 26th. According to public information, “Hug Your Face” is an online platform like an “AI software supermarket.” In mid-July, the company said its system had been invaded by an “autonomous” AI agent. On July 22, OpenAI admitted that during their tests, the two models ran out of the test “cage” themselves, connected to the Internet, and broke into the internal system of “hugging the face.” This is the first time in history that AI “autonomously” attacks other systems.