On August 18, 2026, the nonprofit Guidelight AI Standards, founded by two former OpenAI safety staff, published its first assessment of AI control practices at five frontier companies: Anthropic, Google, Meta, OpenAI, and xAI. The evaluation found that basic practices for retaining control of AI systems are, at most, partially implemented industry-wide, with Anthropic and OpenAI showing the strongest controls and Meta and xAI lagging furthest behind. Prevention and containment measures were identified as the weakest area across all five companies.
