tech ๐ค AI labs hide how they stop rogue models
A recent study found that few top AI labs have published concrete containment response plans for rogue models. Guidelight AI Standards assessed five leading labs, with OpenAI ranking best and Anthropic and Meta scoring lowest. This matters as autonomous AI roles grow inside companies, and regulators in California and New York demand disclosures. The assessment highlighted that companies are less vocal about what happens when operating models misbehave. The study pushes for greater transparency amid growing regulatory pressure, like California's SB 53. ๐ค