xible April 2026 · xachyk@xible.dev

Nobody is scoring AI risk from the outside.

OpenAI scores

0/100
40 55
0100

Nobody was keeping score. Bottom quartile. Below Jasper. Below HireVue. Below Synthesia.

Scored on what they do, not what they say. 22 indicators -- incidents, litigation, regulatory exposure, benchmarks, transparency, financial health. Public data. No cooperation. Built for anyone pricing AI risk.

Higher is safer
55+ low risk 50-54 40-49 elevated <40 high risk
#CompanyScoreSector

Safety benchmarks don't predict safety

Anthropic (#8) and OpenAI (#43) post among the highest safety benchmark scores in the dataset. The gap between them is incidents and litigation.

The auditors are conflicted

The same firms that sell AI advisory audit their clients' governance. No published study links AI certification to fewer AI incidents.

Scale doesn't buy good governance

Microsoft (#34), Amazon (#35), Apple (#38), and OpenAI (#43) all rank in the bottom third despite having the most resources.

English-language bias

Baidu (#1) and Alibaba (#3) reflect lower documented incidents in English-language databases -- not necessarily better governance.

Scores are relative, not absolute. The middle band (40-55) clusters 36 of 50 entities. Differences of a few points may not be meaningful. The extremes are robust.