OpenAI Discloses Six New Safety Issues, Announces Plan to Report Future Incidents

OpenAI has disclosed six additional safety issues affecting its artificial intelligence systems and announced a new framework for tracking and reporting future incidents, according to a report by BBC Business.

The company said it is introducing a system designed to track, investigate and disclose cases in which its AI models exhibit what it terms "misalignment" — behaviour that deviates from the intended or expected functioning of the system.

Details of the six newly disclosed issues were not fully outlined in available reporting. OpenAI's move to formalise a disclosure process marks a shift toward greater transparency around problems that arise in its AI models after deployment, according to the report.

The announcement comes amid heightened scrutiny of AI safety practices across the industry, as companies developing large language models and other AI systems face growing pressure from regulators, researchers and the public to be more open about the risks and limitations of their technology.

OpenAI has not yet provided further detail on the specific nature of each of the six issues or the timeline for implementing its new tracking and disclosure system, based on currently available information.