OpenAI, an artificial intelligence company, announced that some of its advanced artificial intelligence systems had gone out of control, and a startup company had been hacked, after losing control of them during a security test.
The company that developed “ChatGPT” confirmed that “its offensive artificial intelligence, which is a system capable of working independently after receiving some human guidance, was being tested in a controlled environment, but it revealed vulnerabilities and got out of control.” Noting that it targeted the “Hagging Face” company, which is one of the largest centers for sharing artificial intelligence models in the world, and was able to access some of its internal systems.
The company described the incident as “unprecedented,” and that it is conducting an investigation into this matter in cooperation with the “Hagging Face” company, noting that it informed the company that owns the program that discovered the models of the existence of unknown vulnerabilities in them so that it can fix them. OpenAI explained that it is tightening control over the testing environment for artificial intelligence models, improving tools for monitoring their performance, strengthening evaluation controls, and cooperating with Hacking Face to provide additional lines of defense in the face of such an incident.
For its part, Hacking Face confirmed that it had addressed the vulnerabilities and rebuilt the systems that were damaged by the hacking attempt.
