Skip to main content
June 9, 2026Anthropic Launches Claude Fable 5 with Runtime Fallback Safeguards and Mandat...→

Weekly newsletter

AI accountability after the demo is over

Every Tuesday, a short field note on what was approved, who owns the decision, and what evidence survives after deployment.

The Governance Gap
Reset

May 26, 2026

The Governance Gap
Latest

Your Logs Pass the Audit. Survive the Examiner.

The governance lead pulled the agent logs and found them spotless. Every interaction captured, every field change timestamped and attributed. None of it could answer the question an examiner actually asks, which is who authorized the specific business decision the agent made. Comprehensive logging and accountability turn out to be different exhibits, and most systems of record can produce only the first.

Read issue 11 →

May 19, 2026

The Governance Gap

The AI Model Started Failing in November. The Organization Found Out in February.

A fraud model ran for seventy-seven days without the controls that justified its deployment. The detection worked. An URGENT email landed at 8:47 on a Tuesday with the sensitivity rate collapsed to 71. The gap was never detection. It was the distance between that email and the model risk committee that met on Thursday, while Maria's rent came due in four days and Patricia's April retirement stopped being certain.

Read issue 10 →

May 12, 2026

The Governance Gap

Four years later, a court letter explained the 1:50am rejection.

The machine rejected Derek Mobley at 1:50 in the morning, before any recruiter arrived the next day. Four years later, a federal court letter told him a machine may have scored his application before anyone saw his name. The organization that deployed the scoring tool had a vendor contract. No named person inside it had ever written down what the system was permitted to decide.

Read issue 9 →