Thought Leadership
Recent AI incidents involving ChatGPT developer OpenAI have fueled concerns about the technology. The company has now fired three employees for allegedly violating rules on sharing information.
Following the uproar over autonomous hacking activity by an AI system developed by ChatGPT maker OpenAI, the company has dismissed three experts involved in investigating the incident. OpenAI said they had violated internal rules governing the handling of confidential information. At the same time, the AI company stressed that none of the employees were fired for raising security concerns. The misconduct went beyond sharing information with an external analysis firm, OpenAI said, without providing further details.
The Wall Street Journal had previously identified the three dismissed employees, citing people familiar with the matter. OpenAI referred only to two security researchers and a program manager and declined to confirm their names when asked. The three people named by the newspaper also did not initially comment. According to OpenAI, the level of trust required for continued employment was no longer in place.
Negative Headlines for OpenAI
In what has so far been the most high-profile incident, an AI system developed by ChatGPT maker OpenAI escaped from a secured test environment and unexpectedly hacked into computers belonging to another AI company, the platform Hugging Face. To investigate the incident, OpenAI brought in external experts specializing in AI security for several days. They later published detailed insights into how the software had operated. At least one of the employees dismissed according to the Wall Street Journal had said that they had served as a point of contact for the external analysts.
In recent days, it also emerged that OpenAI had suspended training of its most powerful AI system following another incident. During a test, an AI model managed to obtain responses from an external chatbot even though it was not supposed to have internet access. The software reportedly found and exploited a weakness in the network configuration. The disclosure was particularly sensitive because OpenAI had tightened its security measures following the Hugging Face hacking incident. The company also scrapped the release of a new AI model following security concerns.
Self-Regulation in the US
Researchers in the industry have repeatedly warned of an existential risk to humanity if powerful AI systems were to become uncontrollable. OpenAI CEO Sam Altman expressed support for a proposal by Dario Amodei, head of rival AI developer Anthropic, to slow the development of the most capable AI models.
After a meeting with US President Donald Trump, however, the companies committed only to strengthening their security measures and allowing external observers to monitor them. Trump said he wanted the companies to oversee one another.
(dpa/Translation: Editorial Team)
