[UPDATE] EU AI Act's transparency rules and AI-fining powers go live; bank credit and AML rules stay delayed to Dec 2027
AI governance · Took effect 2 Aug 2026; high-risk bank rules now confirmed for 2 Dec 2027
As of 2 August, the EU AI Act's general transparency obligations — chatbot disclosure, deepfake labelling, AI-content marking — took effect as originally scheduled, and the European Commission's power to fine general-purpose-AI providers (up to 3% of global turnover or €15m) activated. This confirms in practice what May's Digital Omnibus agreed: the heavier high-risk regime covering bank credit-scoring and AML systems does not bite until 2 December 2027.
Why it matters — Banks operating or planning AI-driven credit and AML systems with EU exposure get more runway than originally drafted — but general disclosure and deepfake-labelling obligations, and GPAI enforcement risk, are live today, not in 16 months.
Model governance named the top barrier to scaling bank AI, new index finds
AI governance
Wolters Kluwer's H1 2026 AI Risk and Governance Index finds more than a third of financial institutions cite model governance and validation — not technology or budget — as the primary barrier to scaling AI, alongside model drift, synthetic-data integrity and vendor dependency as the risks now defining AI at scale.
Why it matters — This lands squarely on the gap flagged locally: the FMA has already named AI governance in financial advice a 2026/27 priority, and this data suggests the binding constraint on every bank's AI roadmap is governance capacity, not appetite.
EU delays AI Act 'high-risk' rules for bank credit and AML systems by 16 months
AI governance · Provisional deal reached 7 May 2026; formal adoption expected before 2 Aug 2026
EU negotiators have provisionally agreed, via the Digital Omnibus, to push back the AI Act's high-risk compliance deadline for standalone systems — including bank credit-scoring, AML and underwriting models — from 2 August 2026 to 2 December 2027, with product-embedded high-risk systems deferred further to 2 August 2028. Formal publication in the Official Journal is expected before the original deadline lapses.
Why it matters — Banks and fintechs building EU-facing AI credit and compliance tools get significant runway relief, but the shifting timeline complicates global model-governance roadmaps and raises the risk that groups with EU operations miscalibrate compliance sequencing against other regimes moving in parallel (US Treasury FS AI RMF, RBI's draft kill-switch rules, FSB sound practices).
Philippines central bank issues voluntary 'STARS' AI governance principles for banks
AI governance · Memo dated 24 Jun 2026, reported 15 Jul 2026
Bangko Sentral ng Pilipinas has issued non-binding governance principles, branded STARS (sustainability, transparency, accountability, responsibility, security), requiring supervised banks and their outsourced AI vendors to keep a centralised AI system inventory and disclose when AI influences a financial decision.
Why it matters — Adds another Asia-Pacific regulator to the fast-growing patchwork of AI governance regimes banks must reconcile, alongside RBI's kill-switch mandate, MAS's SAFR framework and the FSB's forthcoming global standard.
FMA flags AI governance in financial advice as a 2026/27 regulatory priority
AI governance · Financial Conduct Report published 30 June 2026
New Zealand's Financial Markets Authority named AI use in financial advice, credit underwriting and pricing a core focus in its second annual Financial Conduct Report, telling firms AI outputs must be reliable, explainable and open to challenge while sensitive data stays protected.
Why it matters — Signals the FMA will actively scrutinise how NZ banks and advisers deploy AI-supported advice and lending tools, raising the governance bar just as institutions accelerate rollout.
Regulatory — NZ horizon & global signalshighUpdate
[UPDATE] FSB to finalise AI 'sound practices' standard this year, scaled by bank size
AI governance · Finalisation timeline confirmed 7 Jul 2026
Fed Vice Chair Michelle Bowman, who chairs the FSB's supervisory-cooperation committee, confirmed the FSB's consultation on sound practices for AI in finance will be finalised later this year as a G20 deliverable, with standards proportional to a bank's size and risk profile rather than one-size-fits-all.
Why it matters — Gives RBNZ and FMA a firm timeline and a proportionality principle to anchor any future local AI-governance expectations, rather than an open-ended consultation.
Financial Stability Board opens consultation on 12 sound practices for AI adoption in finance
AI governance · Published 10 June 2026; consultation closes 22 July 2026, final report due October 2026
The FSB published a consultation report proposing 12 sound practices spanning board-level AI governance, lifecycle risk management, and AI-related cyber/third-party risk, explicitly addressing generative and agentic AI.
Why it matters — Gives a global standard-setter benchmark that national regulators, including New Zealand's, are likely to reference when firming up local AI governance expectations, ahead of a hard October 2026 finalisation deadline.