Enterprise AI and Security Coverage Roundup (OpenAI–Pentagon Deal, AI Security Tools, and Governance Commentary)

Citizen Lab highlighted concerns about OpenAI’s Pentagon contract, noting expert skepticism that bulk user-data collection can be effectively ruled out and warning that feeding commercially available personal data into opaque AI systems can amplify harm through errors, bias, and weak accountability. Separately, CSO Online reported on OpenAI’s security-related initiatives, including a claim that Codex Security identified 11,000 “high-impact” bugs in a month and a report that OpenAI plans to acquire Promptfoo to strengthen AI agent security testing.

Most other items in the set are opinion/feature or promotional content rather than incident-driven threat intelligence: CIO and CSO Online ran general enterprise AI and security management pieces (e.g., “shadow AI” governance, identity decisioning, OT/IoT/zero trust challenges, cloud security culture/process issues, and pen-test automation lessons learned), while Red Canary published an RSAC 2026 session guide. One CSO Online headline referenced a critical HPE Aruba CX switch flaw enabling admin control without credentials, but the provided text does not include details sufficient to confirm it as the same story as the OpenAI items and it appears as a sidebar link rather than the primary subject of the referenced pages.