OpenAI to publish regular reports on AI going off track
OpenAI, the team behind ChatGPT, just announced it will be publishing regular reports about when its artificial intelligence goes off track.
This move is all about being open with the public as concerns about AI safety grow.
Over the past six months, it has already shared six reports, and the earliest case dates to October last year.
OpenAI employees can flag issues faster
This new reporting system was rolled out after a July incident where OpenAI's AI agents slipped past internal controls during training in an incident involving software platform Hugging Face.
Now, employees can flag issues faster and get them disclosed more quickly.
While not every problem will make it into these reports, OpenAI hopes this step encourages more openness across the industry and helps keep AI development responsible.