OpenAI disbanded its preparedness team, report says

OpenAI disbanded its preparedness team at the end of last month, according to a report by the Financial Times. The unit was tasked with a specific function: determining if the company’s artificial intelligence models presented serious risks.

Before its dissolution, the team focused on identifying potential dangers associated with advanced software. Their work involved analyzing whether models could behave in unintended or harmful ways. One specific area of concern was the potential for a system to act autonomously, such as attempting to hack another company.

Beyond identifying threats, the group was responsible for creating safeguards. They developed methods to mitigate the risks they found, aiming to prevent dangerous outcomes before the technology reached the public.

The Financial Times reported that responsibility for these duties has shifted. The specific details regarding where the team’s obligations now lie were not fully detailed in the initial reporting.

The preparedness team operated as a checkpoint within the company. It evaluated systems that the company believed possessed capabilities that could outperform humans in certain tasks. This internal evaluation process was designed to ensure safety measures kept pace with rapid development.

Artificial intelligence safety has become a major focus for regulators and technology companies alike. The disbanding of this specific group comes as the industry continues to debate the best ways to manage powerful new tools.