Center for AI Safety (CAIS) — publication: The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning — benchmark for evaluating dual-use AI capabilities in biosecurity, cybersecurity, and chemical weapons

NoteThe source directly confirms all key elements of the claim: (1) CAIS (Center for AI Safety) is listed as an affiliation for multiple authors including lead author Nathaniel Li; (2) the publication title matches exactly; (3) it is described as a benchmark for evaluating dual-use AI capabilities; (4) the three domains mentioned (biosecurity, cybersecurity, and chemical weapons/security) are explicitly confirmed; (5) the authorship by Li et al. is confirmed; (6) the date matches (2403 = March 2024, which falls within 2024-03). All factual claims in the assertion are supported by the source text.