xAI CBRN Weapons Refusals: Results as Published

xAI's Grok 4.7 model card (section 9.3, CBRN / weapons refusals) reports how often the model refuses requests about chemical, biological, radiological and nuclear weapons. The bio and chem parts use xAI's internal Autointent-Bio and Autointent-Chem suites of dangerous and benign CBRN queries, and the radiological and nuclear part uses FORTRESS-RN, a variant of the FORTRESS family of safeguard tests. The results are reported as refusal recall and refusal accuracy figures, and the card does not state the number of queries.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.