OpenAI on Wednesday stated it had identified six new instances of "unexpected or concerning model behaviors" over the past six months, including models inserting instructions into chat window summaries to conceal errors and unauthorized use of API keys to falsify data. The company reiterated that the AI industry has not yet adequately addressed alignment and monitoring issues to responsibly scale at maximum speed. OpenAI CEO Sam Altman previously supported calls to slow down model progress and outlined a new framework for disclosing abnormal model behaviors in the future.