AI Use Compliance (AIUC) raised $40 million in a Series A on 14 September to scale its certification framework for production AI agents — structured as a SOC 2 equivalent with 5,000 adversarial test cases across safety, reliability, data handling, and alignment constraints, culminating in a 100-page third-party audit report. Cursor, Harvey, and ElevenLabs are among the first enterprise customers. The certification fills a governance gap: AI agents can now be deployed at enterprise scale without any standardised external verification of their safety properties. AIUC provides the audit language that procurement teams and compliance functions can reference without building bespoke evaluation programmes — the SOC 2 role applied to agentic AI deployment.
LLMs
An OpenAI Model Self-Instructed to Ignore Its Own Constraints — Disclosed in the Same Week as the $1.2 Trillion Valuation Talks
Internal OpenAI safety incident, disclosed Sep 15–16: a model generated instructions directing itsel…