The six that came due this week

7 November, wrong. The call needed a major bank or insurer outside Australia to publish a named operational AI risk function, with a third-party AI vendor register, a change gate for production updates and a board-level owner, inside its annual risk-and-controls statement. Nothing of that shape appeared. The call also waited on labels we invented rather than words a bank actually prints.

8 November, wrong. It needed a major pharma to publish an OpenBind-trained baseline against its closed-vendor binding model. OpenBind's own trail ends with its May release.

9 November, wrong. It needed a named inventory of AI-generated code inside a 10-K risk factor or a post-breach notice, with the credential-scan cadence and the remediation timeline. No filing carries one.

10 November, wrong. It needed an ERISA class action alleging that a plan committee failed to monitor an AI-driven prior-authorisation or claims-review denial. No such filing surfaced.

11 November, wrong. It needed a Tier-1 pharma to disclose a multi-billion-dollar acquisition of an AI-native discovery platform on an earnings call or in a 10-Q. The closest event in the window runs the other way: Isomorphic Labs is reported to be seeking $5 billion of its own on 8 November.

12 November, wrong. It needed a US state settlement in which the violation cited an AI-driven processing pipeline. State activity in the window is new legislation and chatbot rules, not a privacy settlement with AI named in it.