OpenAI AI Updates: September 22, 2026

1. OpenAI Set Up an External Math Advisory Group After Its Models Cleared 100 Open Problems

OpenAI announced an independent Advisory Group on Mathematics and Artificial Intelligence to review and communicate emerging results, as its systems have now resolved more than 100 open mathematical problems. The group’s remit is verification and framing, not direction: TechCrunch reports it will not have the authority to slow down or redirect the underlying research. The distinction matters because the bottleneck on claims like these has been independent checking, and OpenAI is buying review capacity without giving up control of the program. Source Source

2. OpenAI Published Its Case for Shared Global AI Standards

OpenAI outlined a path toward common international AI standards, calling for coordinated evaluation, reporting, and governance across labs and jurisdictions. The position arrives in the same week the UN science panel warned that agent control is not assured and the US and China agreed to a formal AI dialogue. A lab proposing the evaluation and reporting format is a live question for anyone who will end up subject to it. Source

3. OpenAI Academy Added Role-Specific Learning Paths and Credentials

OpenAI expanded OpenAI Academy with new learning paths aimed at employees, developers, leaders, educators, and students, structured so participants can build and then demonstrate practical AI skills. The addition of a demonstrable credential, rather than course material alone, is the change worth noting for teams standardizing internal AI training. Source