ChatGPT Astra: OpenAI Adds Task Review Checkpoints

ChatGPT Astra is rolling out with new review steps for its AI models. OpenAI says this upgrade is for tasks like coding, research, and complex multi-step work. The company has added a review step for situations where Astra's behavior might need closer inspection. This means Astra can pause or stop a conversation if it detects a possible failure to correctly interpret instructions. Users can then review the situation and decide if the task should continue. OpenAI says Astra can create documents, spreadsheets, and presentations while following user templates. This makes it useful for office work and software tasks. However, a mistake could travel through several steps before someone notices. The new controls aim to address this. For developers, a stopped task might be the model's safety system asking for human input. This matters because it shifts responsibility to the user for the next step in the workflow.