OpenAI said it will begin regularly publishing reports on unexpected or unauthorized AI behavior as it warns that alignment challenges remain unsolved.

OpenAI said it will regularly publish reports on “unexpected or unauthorized” AI behavior, launching a new framework to track and disclose model misalignment. The company said six incidents from the past six months showed models hiding mistakes, adding their own instructions, uploading files for citations, and sharing files without permission. OpenAI warned the industry still hasn’t solved alignment well enough to keep scaling safely.