AgentRisk · Corrections

Corrections — Intelligence Brief 001

v1.0 → v1.1 Published correction date: 19 July 2026 Publication date retained: 16 July 2026

What changed

  1. FSB finding corrected. The Financial Stability Board finding now rests primarily on its 10 June 2026 consultation report, Sound Practices for Responsible Adoption of Artificial Intelligence (AI), which proposes 12 sound practices. The 29 June outreach notice and 7 July opening remarks are now supporting references rather than the finding's basis.
  2. JFSA/BOJ dates corrected. The joint request was issued on 22 May 2026. The 15 June 2026 date is the publication date of the provisional English translation, not the issuance date.
  3. Currency explanation added. The JFSA/BOJ item is retained as older context because the English version made the request accessible to international risk teams. It is no longer presented as a fresh 30-day issuance.
  4. Findings reordered. The FSB consultation and JFSA/BOJ short-term-measures request now appear before the FCA review because they are more immediately relevant to control design and resilience. The FCA review remains important as a longer-horizon strategic signal.
  5. Ratings differentiated. The three editorial assessments now distinguish Elevated governance pressure, High resilience urgency, and Moderate strategic direction. "Composite Threat Level" has been replaced with "Overall assessment."
  6. Publishable wording cleaned. Non-public production terminology was removed. The article now uses public-facing, source-backed language.
  7. Control language made practicable. "Immutable action logging" was replaced with "tamper-evident, retained logging of material AI actions."
  8. Ownership language made adaptable. The article now states that accountable leadership is typically the CRO/CISO, adapted to the institution's structure, rather than prescribing a universal assignment.

What did not change

The publication remains a source-backed interpretation of three official financial-sector materials. It does not assert a single harmonised legal obligation or that every institution has experienced a frontier-AI incident.

Source basis

← Back to Intelligence Brief 001