Regulatory Calendar for AI in EU Financial Services
The stop-the-press calendar rebuilt as a dated register: AI Act stages after the AI Omnibus, DORA and its RTS/ITS, CTPP designations, ESA and supervisor guidance, national laws such as Italy's Law 132/2025 with corrected dates, and harmonised standards, each with status, as-of date, official source and the reference pages it moves, plus the change-trigger procedure for when an item moves.
- Version
- 0.1.0
- Owner
- DEIXEI
- Last reviewed
- Reading time
- 18 min
For Cto · Ai Architect · Risk And Assurance · Compliance · Internal Audit · Legal And Compliance · Programme Office
Sources and details
- Section
- Govern and prove
- Source works
- Book 1 · Book 2 · Book 3 · Book 09 · Research Series · Whitepaper
- Source chapters
- 1, 3, 6, 7, 26
- Source appendices
- 01:C, 04:I
- Refresh trigger
- regulatory change
This page is the factory’s regulatory calendar. It rebuilds Book 1’s Stop-the-Press Calendar (Appendix C, as of 10 May 2026) as a dated register: each item has a status, the date that status was checked, the official source to re-check it against, and the reference pages it moves. It also carries the procedure for what to do when an item moves.
It is the only place in the reference that states regulatory dates. The regulatory map holds the article-level obligations and controls. Every other page links to these two.
As of 14 November 2026; official sources re-checked 15 November 2026. Only dates marked Verified are stated as fact. Everything else is reported as the source book gives it, with the source named, and carries a [FACT-CHECK: …] marker. EUR-Lex refused automated retrieval during this check, so the Commission’s AI Act Service Desk (which publishes the consolidated text) and the Official Journal text hosted elsewhere were used instead.
How to read an entry
| Column | Meaning |
|---|---|
| Date | The application, publication or deadline date. ”—” where no date is set |
| Item | What applies or is expected |
| Status | In force (law applies now); Scheduled (adopted law with a future application date); Pending (expected but not adopted or published); Delayed (a legal or announced deadline passed without delivery); Unverified (source claim not yet checked) |
| Checked | The date the status was last confirmed, and Verified where it was confirmed against an official source |
| Official source | Where to re-check. A book citation is not an official source |
| Moves | The reference pages whose content changes if this item moves |
Composite firms (Bank A and others) do not appear on this page. Vendor product releases (agent platforms, SDKs, model line-ups) are not regulatory items; they are tracked on the vendor pages with a vendor-release refresh trigger, such as deployment routes.
AI Act application stages
Regulation (EU) 2024/1689 as amended by the AI Omnibus. Dates verified against the AI Act Service Desk timeline and Article 113 as consolidated at 27 July 2026, and the Commission’s AI Act policy page (last updated 19 October 2026), all accessed 15 November 2026.
| Date | Item | Status | Checked | Official source | Moves |
|---|---|---|---|---|---|
| 12 July 2024 | AI Act published in the Official Journal (per Book 1 Ch 6–7) | In force | Unverified [FACT-CHECK: OJ date on EUR-Lex] | EUR-Lex | — |
| 1 August 2024 | Entry into force | In force | Verified 15 Nov 2026 | Service Desk timeline; Commission policy page | — |
| 2 February 2025 | General provisions and prohibitions (Chapters I–II): definitions, AI literacy, Art 5 prohibited practices | In force | Verified | Service Desk; Art 113(a) | Risk tiers (prohibited-use list) |
| 2 August 2025 | GPAI model obligations; governance bodies; national authority designation and penalty rules (Chapter III Section 4, Chapters V, VII, XII, Art 78; Art 101 excepted) | In force | Verified | Service Desk; Art 113(b) | Regulatory map supplier rows |
| 27 July 2026 | AI Omnibus in force; Arts 102–110 (amendments to other Union acts) apply from this date | In force | Verified | Commission policy page; Art 113(d) | — |
| 2 August 2026 | General application date; enforcement begins; Art 50 transparency duties apply; innovation-support measures start | In force | Verified | Service Desk timeline; Art 113 | Regulatory map; client-channel pages |
| 2 December 2026 | New prohibitions (Art 5(1) points (ba), (bb) and Art 5(1a), (1b)): AI systems generating non-consensual sexual deepfakes and child sexual abuse material | Scheduled | Verified | Art 113(a) as amended; Service Desk timeline | Risk tiers (prohibited-use list) |
| 2 December 2026 | Transitional deadline: providers of systems placed on the market before 2 August 2026 must comply with Art 50(2) marking, synthetic-content generators included | Scheduled | Verified | Service Desk timeline | Regulatory map |
| 2 August 2027 | Each Member State to have at least one AI regulatory sandbox operational | Scheduled | Verified | Service Desk timeline | — |
| 2 August 2027 | GPAI models placed on the market before 2 August 2025 must comply (per Book 3 App I) | Scheduled | Unverified [FACT-CHECK: Art 111(3) as consolidated] | Service Desk, Art 111 | Model governance |
| 2 December 2027 | High-risk rules (Chapter III Sections 1–3) for Annex III systems under Art 6(2) | Scheduled | Verified | Art 113(c)(i) as amended | Regulatory map (Arts 9–15, 26, 27, 49); risk tiers; audit evidence |
| 2 August 2028 | High-risk rules for Annex I product-embedded systems under Art 6(1) | Scheduled | Verified | Art 113(c)(ii) as amended | Same |
Planning posture. The deferral moved dates, not obligations. Book 3 App I’s recommendation, kept on the regulatory map, is to build Tier 3 controls to the original baseline and arrive at 2 December 2027 with 12–18 months of operating evidence.
The AI Omnibus itself
| Fact | Value | Checked |
|---|---|---|
| Proposed | 19 November 2025 | Verified (Commission policy page) |
| Political agreement | 7 May 2026 | Verified (Commission policy page) |
| Amending act | Regulation (EU) 2026/1744 (the Commission page links OJ L 2026/1744) | Verified as the OJ reference; regulation number as cited by Book 2 Ch 26, Book 09 Ch 1 and Research Paper 1 |
| Date of the act | 8 July 2026 | Unverified [FACT-CHECK: date on the OJ text] |
| Published | OJ L, 24 July 2026 | Unverified [FACT-CHECK: OJ publication date] |
| Entry into force | 27 July 2026 | Verified (Commission policy page) |
| Other changes reported | Art 4 replaced: providers and deployers take measures to support AI literacy of their staff (Research Paper 1, Research Paper 9); stricter necessity test for special-category data in bias detection; SME relaxations extended to small mid-caps; competence clarification keeping national authorities in the lead for financial institutions (Book 1 Ch 7, describing the political agreement) | Unverified [FACT-CHECK: confirm each against the adopted text] |
Pending AI Act items
| Date | Item | Status | Checked | Official source | Moves |
|---|---|---|---|---|---|
| Was due 2 February 2026 | Commission guidelines on high-risk classification (Art 6(5)); Book 1 records the deadline missed and a draft in consultation | Delayed | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK: publication status at 15 Nov 2026] | Commission (DG CNECT); Service Desk | Risk tiers; every sectoral position paper (step 1 of the diagnostic) |
| — | Harmonised standards for high-risk requirements (CEN-CENELEC, standardisation request M/593). prEN 18286 (quality management systems) entered public enquiry 30 October 2025; first harmonised standards expected end 2026 (Book 1 App C, Ch 6–7) | Delayed | As of 10 May 2026 (Book 1 App C) [FACT-CHECK: CEN-CENELEC status and any OJ citation] | CEN-CENELEC; Official Journal (citation of references) | Control map (ISO/IEC 42001 rows); regulatory map QMS row |
| — | Codes of practice for detecting and labelling AI-generated content (Art 50(7), amended by the Omnibus) | Pending | Amendment marker verified 15 Nov 2026; content and date unverified [FACT-CHECK] | Service Desk, Art 50 | Regulatory map Art 50 row |
| Was due 2 August 2025 | Commission guidance on Art 73 serious-incident reporting (Art 73(7)) | Unverified [FACT-CHECK: whether issued] | Deadline verified (Art 73 text) | Service Desk, Art 73 | Incident response |
| — | Scope and start of the Art 86 right to explanation for Annex III decisions after the amended Art 113 | Unverified [FACT-CHECK] | — | Service Desk, Art 86, Art 113 | Regulatory map subject-request section |
DORA and its RTS/ITS
| Date | Item | Status | Checked | Official source | Moves |
|---|---|---|---|---|---|
| 17 January 2025 | DORA (Regulation (EU) 2022/2554) applies (Art 64) | In force | Verified (EIOPA DORA page) | EUR-Lex; EIOPA | All DORA rows on the regulatory map |
| 25 June 2024 (OJ) | RTS 2024/1774 on the ICT risk-management framework, including Art 17 change management | In force | OJ date as cited by Book 2 Ch 13 and Book 09 [FACT-CHECK: OJ date] | EUR-Lex | Change governance; regulatory map change and logging rows |
| — | RTS 2024/1772 on incident classification (Art 8 major-incident test) | In force | Unverified [FACT-CHECK: OJ reference and date, as Book 09 Ch 8 also flags] | EUR-Lex | Incident response |
| 20 February 2025 (OJ) | RTS 2025/301 on content and time limits for major-incident notifications and reports (4 h / 24 h / 72 h / one month) | In force | Verified (OJ text) | EUR-Lex ELI reg_del/2025/301 | Incident response; regulatory map clocks |
| 20 February 2025 (OJ) | ITS 2025/302 on incident-reporting templates (per Book 1 Ch 6) | In force | Unverified [FACT-CHECK] | EUR-Lex | Incident response |
| — | RTS 2024/1773 on the policy for ICT services supporting CIFs from third parties | In force | Unverified [FACT-CHECK] | EUR-Lex | Vendor concentration and exit |
| — | ITS 2024/2956 on Register of Information templates (adopted 29 November 2024, per Book 1 Ch 6) | In force | Unverified [FACT-CHECK] | EUR-Lex | Regulatory map RoI section |
| Annual (window set by competent authorities) | Register of Information submission (Book 3 App I: “typically Q1”) | Recurring | Unverified [FACT-CHECK: this year’s window from the firm’s competent authority] | National competent authority; ESAs | Regulatory map RoI section; vendor concentration and exit |
| — | RTS 2025/532 on subcontracting (17 April 2025 per Book 1 Ch 6) | In force | Unverified [FACT-CHECK] | EUR-Lex | Supplier rows |
| 8 July 2025 | RTS 2025/1190 on threat-led penetration testing applies (adopted 13 February 2025, OJ 18 June 2025, per Book 1 Ch 6) | In force | Unverified [FACT-CHECK: all three dates] | EUR-Lex | Evaluation and resilience |
| Before 17 January 2028 | First mandatory TLPT exercises to complete (Book 1 Ch 6 and App C) | Scheduled | Unverified [FACT-CHECK: the legal basis of this deadline] | TLPT authority per Member State | Evaluation and resilience; regulatory map TLPT section |
CTPP designations
| Date | Item | Status | Checked | Official source | Moves |
|---|---|---|---|---|---|
| 18 February 2025 | ESAs roadmap for CTPP designation published (dated link) | Published | Verified (listed on EIOPA DORA page) | EIOPA | — |
| 18 November 2025 | First list of designated critical ICT third-party providers published by the ESAs | Published | Verified (ESMA press release) | ESMA, EBA, EIOPA | Regulatory map supplier posture; deployment routes; vendor concentration and exit |
| 18 November 2025 | Content of that list: nineteen providers including AWS, Microsoft and Google Cloud; neither Anthropic nor OpenAI (Book 3 Ch 3 §4, App I §3, as of August 2026) | Published | Unverified [FACT-CHECK: read the list PDF] | ESAs list PDF | Same |
| — | Next designation round; any designation of a model provider | Pending | Book 3 Ch 3 records none published as at its writing [FACT-CHECK: status at 15 Nov 2026] | ESAs Joint Committee | Same. A designation of Anthropic triggers mandatory re-versioning of the DORA rows (Book 3 App I §11) |
| Late 2026 | First public CTPP oversight findings expected (Book 1 App C) | Pending | Unverified [FACT-CHECK] | ESAs | Vendor concentration and exit |
ESA and supervisory guidance
None of these are binding rules. They change supervisory expectations, which is why the regulatory map labels them Guidance.
| Date | Item | Status | Checked | Official source | Moves |
|---|---|---|---|---|---|
| 30 May 2024 | ESMA statement on AI in retail investment services (MiFID II binds regardless of AI) | Published | As cited by Book 3 App I [FACT-CHECK] | ESMA | Regulatory map sectoral table |
| 6 August 2025 | EIOPA Opinion on AI governance and risk management (EIOPA-BoS-25-360) | Published | Verified (EIOPA) | EIOPA | Regulatory map sectoral layer |
| 28 July 2025 | ECB Guide to Internal Models, revised (ML expectations) | Published | As cited by Book 3 App I [FACT-CHECK] | ECB Banking Supervision | Model governance |
| 21 November 2025 | EBA factsheet on AI Act implications for the banking sector | Published | Date verified via EBA listing; title differs between sources (see regulatory map) | EBA | Regulatory map sectoral layer; Art 26 crosswalk |
| 18 December 2024 | FINMA Guidance 08/2024 on governance and risk management when using AI | Published | As cited by Book 3 App I and Research Paper 9 [FACT-CHECK] | FINMA | Regulatory map Switzerland table |
| 4 November 2025 | Swiss Bankers Association Cloud Guidelines, 3rd edition | Published | As cited by Book 3 App I [FACT-CHECK] | SBA | Same |
| 18 December 2025 | BaFin guidance on ICT risks in the use of AI at financial entities (non-mandatory) | Published | As cited by Book 3 App I [FACT-CHECK] | BaFin | Regulatory map sectoral layer |
| 26 February 2026 | ESMA supervisory briefing on algorithmic trading (RTS 6 self-assessment covers AI/ML) | Published | As cited by Book 3 App I [FACT-CHECK] | ESMA | Regulatory map sectoral table |
| 2026–2027 | EBA follow-up work on AI (credit risk, consumer protection, integration with EBA Guidelines) | Pending | As of 10 May 2026 (Book 1 App C) [FACT-CHECK] | EBA publications | Sectoral layer |
| 2026 | EIOPA follow-up review of its AI Opinion’s implementation | Pending | As of 10 May 2026 (Book 1 App C) [FACT-CHECK] | EIOPA publications | Sectoral layer |
| — | ESMA sector mapping comparable to the EBA’s for asset management | Pending | As of May 2026 (Book 1 Ch 6–7) [FACT-CHECK] | ESMA | Sectoral layer |
National laws
Italy, Law 132/2025 (corrected dates)
| Date | Item | Status | Checked | Official source |
|---|---|---|---|---|
| 23 September 2025 | Legge 23 settembre 2025, n. 132, “Disposizioni e deleghe al Governo in materia di intelligenza artificiale” | Adopted | Verified (Gazzetta Ufficiale) | gazzettaufficiale.it |
| 25 September 2025 | Published in Gazzetta Ufficiale, Serie Generale n. 223 | Published | Verified | Same |
| 10 October 2025 | Entry into force | In force | Verified | Same |
| By 10 October 2026 | Legislative decrees under Art 24 delegations (banking, financial services, insurance, payments, public administration, healthcare, labour), due within 12 months of entry into force per Book 1 Ch 6–7 | Unverified [FACT-CHECK: which decrees were adopted, and whether the delegation period was met or extended] | — | Gazzetta Ufficiale; Normattiva |
| — | Implementing acts on procurement, public administration and the authorities’ operational competence (Book 1 App C) | Pending [FACT-CHECK] | As of 10 May 2026 | Gazzetta Ufficiale; AgID; ACN |
Correction. Book 1 App C says the law was adopted on, and in force from, 17 September 2025. That is wrong on both counts. The fact-checked Book 1 Ch 6–7 and the Gazzetta Ufficiale give the dates above. The calendar follows the Gazzetta.
What it does, per Book 1 Ch 6–7 (Law, as described; confirm against the text). Designates AgID (notifying authority) and ACN (supervisory and sanctioning authority); preserves Banca d’Italia, Consob and IVASS as sectoral supervisors; adds criminal-law overlays, including AI-assisted market manipulation under the TUF; steers public procurement towards Italian data storage. It does not add obligations beyond the AI Act. A firm with Italian operations tags affected agents for review when each sectoral decree is published.
Other Member States and the UK
| Jurisdiction | Item | Status as the source gives it | Checked | Official source |
|---|---|---|---|---|
| Germany | KI-MIG, a narrower implementing act designating BNetzA as market-surveillance authority; Cabinet draft 10 February 2026 | Pending | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK] | Bundesgesetzblatt |
| Netherlands | Uitvoeringswet AI-verordening; consultation 20 April–1 June 2026; eight sectoral authorities | Pending | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK] | Staatscourant; overheid.nl |
| Spain | Anteproyecto on good use of AI (closest to the Italian model); AESIA already designated | Pending | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK] | BOE |
| France | Activity centred on a Senate bill on AI and cultural content, not a framework law | Pending | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK] | Légifrance |
| Belgium | No framework bill; BIPT/IBPT designated as market-surveillance authority in the 2025–2029 government agreement | Pending | As of 30 May 2026 (Book 1 Ch 7) [FACT-CHECK] | Moniteur belge |
| United Kingdom | UK GDPR Art 22 substituted by the Data (Use and Access) Act 2025 (Research Paper 2) | Adopted | As of 10 November 2026 (Research Paper 2) [FACT-CHECK: commencement date] | legislation.gov.uk |
Book 1 App C’s earlier framing (“Italy’s law is the template”) is superseded by Ch 7: other Member States are mostly passing narrower implementing acts, not framework laws. Keep a national-overlay tracker per Member State of operation (see the regulatory map).
Standards and threat catalogues
Not law, but App C tracks them because the control map cites them. They move on their own cadence.
| Date | Item | Status | Checked | Source | Moves |
|---|---|---|---|---|---|
| 9 December 2025 | OWASP Top 10 for Agentic Applications v1.0 (ASI01–ASI10); a v2.0 cadence expected late 2026 (Book 1 App C) | Published / Pending | As cited [FACT-CHECK] | genai.owasp.org | Evaluation and resilience; control map |
| November 2025 | MITRE ATLAS v5.1.0 (Book 3 App I) | Published | As cited [FACT-CHECK] | atlas.mitre.org | TLPT scenario library |
| — | ISO/IEC 42001:2023 stays the QMS reference until harmonised standards are cited | In force (voluntary) | — | ISO | Control map |
When an item moves
Book 1 App C’s discipline is three questions, asked every time an item moves. This page turns it into a procedure with an owner and a record.
Owner (Advice). Legal and Compliance owns the calendar. The regulatory-map owner and the programme office are consulted on every move. Internal Audit reads the change log quarterly.
Watch cadence (Advice). Check the Service Desk timeline, the ESAs’ press pages and the national gazettes of each Member State of operation at least monthly, and on any announced adoption, publication or designation. Book 1 App C suggests checking the OWASP tracker at least quarterly.
The procedure
- Confirm it moved. A move is a publication in the Official Journal or a national gazette, an entry into force, an official designation list, or an authority’s own publication. A press report, a political agreement or a draft is not. Record the official source. Until the canonical text is there, the previous version stands (Book 1 App C).
- Ask the three questions.
- Does it change a citation on the regulatory map or the control map? If so, update the row and re-date the page.
- Does it change a deadline on the firm’s programme plan? If so, the programme plan is amended within the week (Book 1 App C).
- Does it change a Monday-morning answer — what a team should do this week on a reference page? If so, update that page’s practical guidance and its
updateddate.
- Find the pages. Use the “Moves” column. If a row has no pages listed and the move changes guidance, add them.
- Re-version dependent artefacts. Book 3 App I §11 names triggers that force re-versioning of the firm’s own regulatory mapping: final AI Act high-risk dates; a CTPP designation of the model provider; new ESA guidance naming AI; new ECB or national-supervisor guidance; new ISO, NIST, OWASP or ATLAS versions; a change in the provider’s Code of Practice adherence; material change in a non-EU regime applied by analogy.
- Communicate carefully. Do not communicate a replanned date externally before the amending text is published and in force (Book 1 Ch 7).
- Log the move. One entry per move, fields below.
Change-log entry
| Field | Content |
|---|---|
item | Calendar row affected |
moved_on | Date of the official event |
source_url | Official source |
old_status / new_status | For example Pending to Scheduled |
q1_map_citation | yes/no, with rows changed |
q2_programme_deadline | yes/no, with plan item and date amended |
q3_monday_answer | yes/no, with pages updated |
pages_updated | Page IDs and new updated dates |
artefacts_reversioned | Firm artefacts (position papers, crosswalks, RoI notes) |
owner / reviewed_by | Legal and Compliance; regulatory-map owner |
Most moves change one or two of the three; the AI Omnibus changed all three (Book 1 App C).
Acceptance checks
- Every row has a status, a checked date and an official source; no row states an unverified date without a marker.
- No reference page other than this one and the regulatory map states a regulatory date. A search of the reference for a date in this calendar finds only links.
- Every change-log entry since the last review answers all three questions.
- After an Official Journal publication, the calendar and the map were updated within a week, and the programme plan within the same week.
Failure modes
- Replanning to a political agreement. Book 3 App I and Book 1 Ch 7 both hold the original dates until formal adoption. The Omnibus was agreed on 7 May 2026 and in force on 27 July 2026; programmes that moved in May moved early.
- Propagating an uncorrected manuscript. Book 1 Ch 6, Book 1 App C and the Whitepaper (Part II §1 and its glossary entry) say the Omnibus accelerated Art 50 watermarking to 2 December 2026. The verified position is above: Art 50 applies from 2 August 2026, and 2 December 2026 is a transitional deadline for 50(2) systems already on the market. Book 1 App C also misdates Italy’s law. The calendar follows the official sources; the manuscripts carry a source-fix.
- Treating the calendar as a to-do list. It is a checking instrument (Book 1 App C). The work is done on the pages it points to.
- Dates copied into local policies. They go stale without a trigger. Link to this page.
Provenance and refresh
Verified on 15 November 2026: AI Act stages and the Omnibus entry into force (AI Act Service Desk timeline and Art 113 consolidated at 27 July 2026; Commission AI Act policy page); RTS 2025/301 OJ date and time limits (Official Journal text); DORA application date (EIOPA); CTPP list publication date (ESMA); EIOPA Opinion date (EIOPA); Italian Law 132/2025 dates (Gazzetta Ufficiale). Everything else is reported as the named source gives it, with its date, and marked.
The calendar items and the three-question discipline come from Book 1 Appendix C (as of 10 May 2026), restructured and updated; corrections come from Book 1 Ch 7’s fact-check and the official sources. Vendor product items from App C (Microsoft Purview SDK, Agent 365, AWS AgentCore, Google’s agent platform) are left to the vendor pages.
Refresh on every official event in the table, and at least monthly while the high-risk deferral is in transition. Re-date the “as of” line at every revision.
This page remains a draft. It needs independent legal review and EUR-Lex verification of every date not marked Verified before it can move to review status.
Source material
- Book 1 (Building Organizations for the Age of AI), Appendix C, The Stop-the-Press Calendar (as of 10 May 2026; the calendar items and the three-question discipline)
- Book 1 (Building Organizations for the Age of AI), Chapter 6, DORA, the EU AI Act, and the Sectoral Layer (as of 15 May 2026; §4.1–4.6)
- Book 1 (Building Organizations for the Age of AI), Chapter 7, The EU AI Act, Sectoral Guidance, and the National Layer (Draft 4, as of 30 May 2026, after fact-check; §4.1, §4.5)
- Book 2 (The AI Delivery Factory), Chapter 26, The Agent Channel (§4 Article 50 and the AI Omnibus)
- Book 3 (Deploying Claude Safely), Chapter 3, The Control Plane Is Yours (§4 CTPP status)
- Book 3 (Deploying Claude Safely), Appendix I, Regulatory Mapping (as of 14 May 2026; §1–§4, §11 maintenance triggers)
- Book 09 (Everything Is Green), Introduction and Chapter 1 (RTS 2024/1774 citation; Regulation (EU) 2026/1744 citation)
- Research Paper 1, Taxonomy for a Reuse-Driven AI Factory (§2 regulatory status after the AI Omnibus)
- Research Paper 9, Data Classes, Processing Lanes and Controller/Processor Roles for AI (§7.5 dates; §9 regulatory map)
- Whitepaper, Building Organisations for the Age of AI (Part II §1; superseded Article 50 framing)
- European Commission, AI Act Service Desk, Timeline for the implementation of the EU AI Act, and Article 113 (consolidated as at 27 July 2026), accessed 15 November 2026
- European Commission, AI Act policy page (digital-strategy.ec.europa.eu), last updated 19 October 2026, accessed 15 November 2026
- Commission Delegated Regulation (EU) 2025/301 (OJ L, 20.2.2025), Official Journal text, read 15 November 2026
- EIOPA, Digital Operational Resilience Act page, accessed 15 November 2026
- ESMA, European Supervisory Authorities designate critical ICT third-party providers (18 November 2025), accessed 15 November 2026
- EIOPA, Opinion on Artificial Intelligence governance and risk management (6 August 2025), accessed 15 November 2026
- Gazzetta Ufficiale della Repubblica Italiana, Legge 23 settembre 2025, n. 132, Serie Generale n. 223 of 25 September 2025, accessed 15 November 2026
Related guidance
-
Route Regulatory Questions to Obligations and EvidenceA dated, article-level reference that routes AI design questions to the EU AI Act, DORA and its RTS, GDPR and sectoral rules, and turns each obligation into a control, an owner and evidence: change records under RTS 2024/1774 Art 17, the 4 h / 72 h / one-month incident clocks, the CIF hinge and its five artefacts, the three-count Register of Information reconciliation, Art 14(4) oversight controls, the Art 26(5) discharge, the FRIA gate, Art 49 and Art 50, SCHUFA-grade review and supplier controls. -
Build a Unified Control MapJoin the firm's AI and technology controls in one traceable operating view: rows are operating concerns, columns are frameworks read at their own altitude, every cell points to live evidence, and existing ITIL, COBIT and CIS artefacts are extended to agents with a five-question test instead of a parallel AI governance track. -
Vendor Concentration and ExitMeasure concentration on three axes against 70% and 85% discussion triggers, keep a tested successor for every material model usage, record eight contract-cleavage points per route, prove evidence portability with a 15-item checklist, and rehearse exit on one material use case every year with a filed close-out report. -
Model Governance: Usage Registry and Upgrade LifecycleGovern models at the grain regulators ask about: a catalogue for what exists, an 18-field usage registry for what is used, risk classified per usage, a registry the policy decision point reads by usage_id, pinned model IDs, and a four-phase upgrade lifecycle from Day −14 to Day +90 with numeric rollback thresholds. -
The Audit Spine and the Evidence RecordHow to capture, join, protect, retain and prove per-action evidence: the AES v0.1.0 record contract, three capture points and one correlation ID, tiered immutable storage, retention and legal hold, replay rehearsal, the route-approval evidence gate, and the Claude-specific ways evidence goes missing. -
Classify Use Cases into Risk TiersThe T0–T4 use-case taxonomy as a working decision reference: the prohibited-use gate, four intake questions, data and action boundaries per tier, the twelve-area control matrix, retention and payload rules, runtime enforcement, approval gates, regulatory routing and re-tiering. -
Change Governance for Agent-Enabled DeliveryRoute agent-speed change by risk without losing the approver: control values that always reach a person, a routing gate that reads the right signals, autonomy earned per change family, a Virtual CAB that prepares a complete dossier, and three bodies with defined escalation. -
Respond to Agent IncidentsSort an agent failure into one of four incident shapes, suspend it in three verified layers, keep humans on every classification and disclosure decision, route one playbook to every authority that needs a cover sheet, and close each incident with a named control change. -
Evaluate, Assure and Test Agent ResilienceDefine a versioned evaluation contract, test the trajectory and the execution boundaries, run three assurance layers into one quarterly pack, map the OWASP Agentic Top 10 to controls that demonstrably fire, bring agents into threat-led penetration testing with a five-section scope, keep a continuous cadence between formal tests, land every finding as a live control-plane artefact, and engineer recovery down to a minimal AI service level on models you host. -
Choose and Contract Claude Deployment RoutesDecide, per use-case class, which route carries Claude into production (Anthropic API, Claude Platform on AWS, Bedrock, Google Cloud, Microsoft Foundry, plus the Microsoft 365 Copilot surface); record who operates, bills and processes; size the diligence to the processor; check residency, zero data retention and HIPAA per feature; sign a contracting matrix row; and run Agent SDK or hosted agents with the controls the product does not supply. Provider-specific and dated. -
Data Classes, Processing Lanes and TokenisationClassify every model request by its data, route it only to a processing lane the class permits, keep derived artefacts inside their source class, and treat retention, roles and tokenisation as recorded, testable decisions.