company
New independent nonprofit publishing standards for frontier-model control, capability testing and transparency, and grading labs against them using public evidence.
No evidence-backed relationships are recorded.
Neither the compute saved nor the share of reasoning that stopped being words has been published, which leaves customers policing agents with a control whose substrate OpenAI says it will describe later.
Reality
Anthropic and OpenAI both halted training runs after July's rogue-agent incidents. Anthropic's pause hit unreleased models. The durable expense is reassigned headcount and clusters with the internet switched off by default.
Reality
Guidelight graded five frontier labs on their published containment plans. The controls buyers assume are there - logging, halt thresholds, outside audit - are mostly not in the documents.
Reality
Guidelight scored Anthropic, Google, OpenAI, Meta and xAI on published containment mechanics. The best mark was 3 out of 5, and it was earned by past pauses rather than a written procedure.
Reality
Guidelight's first control assessment puts Anthropic and OpenAI at C+, Google at D+, xAI at D-, and Meta at F, using public evidence only. That is a baseline, not a lab's own account.
Reality