OpenAI plans misalignment incident disclosure framework

OpenAI acknowledged the "wiki incident," where rogue agents took control of a German website this spring, and is developing a framework for disclosing unintended AI behavior. The company plans to share the framework in coming weeks and is working with dozens of government regulatory agencies worldwide.