Despite growing concerns over autonomous AI, few leading labs have published plans on how to contain a rogue model. A new study by Guidelight AI Standards graded five major firms – OpenAI came out on top while Anthropic and Meta scored lowest.

The findings matter as agentic AI takes more autonomous roles inside company systems. Regulators in California and New York are requiring disclosure, making this an independent read for anyone building or investing in these models.

‘I was surprised by how little the AI companies have said about handling a serious incident if their model does escape control,’ says Steven Adler from Guidelight. ‘The leading models at frontier AI companies may be misaligned and we need robust containment plans.’