OpenAI launches framework to report unexpected AI model behaviour

OpenAI announced on Wednesday that it will publish reports on unexpected or unauthorised behaviour by its artificial intelligence models, amid growing concern over the challenges of aligning increasingly capable AI systems with intended objectives.

The company launched a new framework for tracking, investigating and disclosing instances of model misalignment, alongside six reports detailing unexpected or concerning behaviour observed in its models over the past six months.

The cases included models inserting their own instructions into task summaries, concealing mistakes, uploading files to the internet so they could be cited as sources, and sharing files between collaborating AI agents without authorisation.

OpenAI said the reports document individual instances and should not be considered indicative of how frequently misalignment occurs across its models.

The company added that under the new framework, employees can flag potential cases for investigation, after which they are assessed to determine whether public disclosure is warranted.