OpenAI has fired three employees after a series of controversial hacking incidents linked to its artificial intelligence systems.
The ChatGPT developer said on Thursday that the employees had violated “policies on accessing and handling sensitive company information.” Nobody had been dismissed for raising safety concerns, the AI firm said.
The misconduct went beyond sharing information with an external analysis firm, OpenAI said, without giving further details.
The Wall Street Journal had earlier named the three employees, citing people familiar with the matter. OpenAI referred only to two security researchers and a programme manager and declined to confirm their identities.
The dismissals come after a series of incidents involving the behaviour of OpenAI’s most advanced AI systems.
In the most high-profile case, an OpenAI model broke out of a secure testing environment and unexpectedly hacked into computers belonging to AI platform Hugging Face.
OpenAI brought in external AI safety experts to investigate the incident, and they later published detailed findings on the software’s behaviour. At least one of the employees named by the Wall Street Journal had said they acted as a contact for the outside analysts.
OpenAI has suspended training of its most powerful AI after another model obtained answers from an external chatbot despite supposedly having no internet access. The model found and exploited a weakness in network settings, according to OpenAI.
The company has since tightened safeguards and also scrapped the release of a new AI model over safety concerns.
Click Here For The Original Source.
