White House AI Safety Framework: Public Excluded

The White House has developed a new framework for evaluating AI models, but the public will not be allowed to see it. This framework is meant to ensure that new AI models don't pose a threat before public release. The White House plans to keep its rubric confidential, sharing it only with the companies whose models will be judged. A few industry players, including those from Anthropic, Google, Meta, and OpenAI, have seen the framework. What's interesting is that the framework is voluntary. AI companies have chosen to comply so far, but could stop submitting their models for evaluation at any time. The initial executive order for this framework requires the government to develop a classified benchmarking process for AI models' cyber capabilities. The bottom line is that a critical safety standard for AI is being kept from the public, making it hard to understand what safeguards are truly in place.