Amazon (arXiv), 2025-07; retrieved 2026-09-15 tier 4 press (primary)confirmed

Evaluating the Critical Risks of Amazon's Nova Premier under the Frontier Model Safety Framework (arXiv 2507.06260, HTML)

Amazon's own report on Nova Premier: Nemesys Insights named as independent CBRN auditor; Amazon shares prompts and outputs with third-party assessors for a final assessment against Amazon-defined thresholds; 120 uplift-indicator prompts. The evaluated lab's document, not the evaluator's; typed press primary for want of a lab-document type. Fetched 2026-09-15; spans present.

Signals citing this source

Nemesys Insights, F (against): Paid assessor for Amazon per the Nova papers; private LLC with no funding disclosed.“We share input prompts and corresponding model outputs with third-party assessors (e.g., Nemesys, METR)”
Nemesys Insights, A (against): Access through client engagement.“Nemesys Insights executed an Exploratory Critical-Capability Threshold Analysis”
Nemesys Insights, S (against): Scope set by the lab client.“their final assessment of whether the model is safe and remains within the defined thresholds for public release”
Nemesys Insights, R (against): Findings appear inside the lab's papers.“reviewed results on test sets, scoring rubrics, and safety-judge notes to verify our internal findings”
Nemesys Insights, M (against): Methods described in the lab's papers only.“They selected 120 "uplift indicator" prompts—60 in synthetic biology and 60 in microbiology/delivery systems”

Ledger rows citing this URL

none