Bottom line: OpenAI's new safety layer can detect misuse without retaining enterprise data, addressing potential risks associated with its AI systems.
What's happening: Researchers at OpenAI have developed a new safety capability that can analyze multiple interactions across various applications, including the LLaMA and Wave models, without retaining prompts or responses, building on the company's existing Zero Data Retention (ZDR) framework.
What to do: Security leaders should monitor OpenAI's API usage and configure their systems to utilize the new safety layer, ensuring risk monitoring while preserving enterprise data compliance.