The patch model is breaking. AI evaluation needs a new way to disclose what it finds. - MLCommons
Traditional coordinated vulnerability disclosure assumes systems can be patched. Here is why that breaks down for AI, and how MLCommons is building a new standard.
MLCommons ยท MLCommons