Richard Barron · Zenodo (CERN European Organization for Nuclear Research) 2026 · 2026
DOI: 10.5281/zenodo.23018092
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
We report the first integrated planning and assessment run of the SPECTER Constellation — three coordinated AI agents operating against a hardened, enterprise-representative agentic AI deployment owned and operated by Red Specter Security Research Ltd. SPECTER AGENT RED generated a four-phase attack sequence selecting tools from the NIGHTFALL v3.0 inventory. SPECTER AGENT BLUE assessed defensive coverage across 14 deployment-specific surfaces. SPECTER AGENT WHITE assessed regulatory compliance across four frameworks. The target scored 9.1% regulatory compliance against ISO 42001, NIST AI RMF, ISO 23894, and ISO 27001 AI. Three unsigned evidence packages produced: SAR-7a4a25f702b8, SAB-7d59e258d04d, SAW-905bb3e5479d. SHA-256 hashes (28 September 2026):SAR-7a4a25f702b8: 61e23289e93b027c2b1fda3a75dfed1c435d97284ca78c75b1b849ef11c28d0aSAB-7d59e258d04d: 0ad1fee03998795563fe7e6f73c25dc9b01597a71a923915274d9d0a23a2e691SAW-905bb3e5479d: d5092d57134d1bf8ccd513f3e4d7d0c5d9dd113d214948501d187f72b25cf641 Verify against the evidence package files uploaded alongside this paper to confirm integrity.
No comments yet — start the discussion below.