AI Compliance Calendar

Checked the Artificial Intelligence Bill, the Data Protection Act 2019 and the ODPC. NO enacted AI law and no binding AI instrument with a stated date. The Artificial Intelligence Bill, 2026 (Senate Bills No. 4 of 2026, sponsored by Senator Karen Nyamu) would create an Office of the Artificial Intelligence Commissioner and a risk-based regime with stringent governance, transparency, data protection and record-keeping duties for high-risk AI, plus disclosure duties on ALL providers and deployers covering 'the nature, purpose and limitations of the system, the extent of automated decision-making, and the measures taken to mitigate biases'. It is NOT LAW: it was read a First Time in the Senate on 2 April 2026 and, because it affects county governments, must also pass the National Assembly before presidential assent. As of reporting through mid-2026 it had not completed passage, and any commencement date would come from a provision that does not yet exist. Kenya's National AI Strategy 2025-2030 is a strategy document, not a duty. The Data Protection Act 2019 contains a general automated-decision-making provision, but it is a general privacy statute whose operative text does not name AI, and its commencement is a 2019/2020 privacy date, not an AI milestone. The Data Protection (Amendment) Bill 2025 proposing AI-related obligations is likewise still a bill.

Watching: Passage of the Artificial Intelligence Bill 2026 by both Houses and presidential assent, followed by a commencement notice in the Kenya Gazette — that would give a real date. Also: an ODPC guidance note or enforcement directive on AI carrying a compliance deadline.

Checked the Information Regulator, POPIA and the national AI policy. NO binding AI instrument with a stated date. (1) South Africa has no standalone AI Act and none is imminent: the Draft National Artificial Intelligence Policy was published by the Department of Communications and Digital Technologies on 10 April 2026 as Notice 3880 of 2026 in Government Gazette No. 54477 — it is a DRAFT POLICY out for comment, not law. Government confirmed the national AI policy will not be finalised until the 2026-2027 financial year and that, when it lands, it will not be standalone legislation but a sector-specific, risk-based approach layered onto existing laws. (2) POPIA section 71 DOES bind and DOES govern automated decision-making — it restricts decisions based solely on automated processing of personal information that result in legal consequences or substantially affect the data subject, and it bites wherever AI affects people in credit, employment or insurance. But its date is a general privacy commencement, not an AI milestone: the substantive sections of POPIA commenced on 1 July 2020 under Proclamation No. R. 21 of 2020 (Government Gazette No. 43461, 22 June 2020) with a one-year grace period, so compliance has been required since 1 July 2021. Section 71's text does not name AI or machine learning. (3) The Information Regulator has issued NO binding AI guidance note with a compliance date; amended POPIA/PAIA regulations took effect 17 April 2025 but concern enforcement machinery, not AI. If the calendar wants a general automated-decision row for South Africa, POPIA s.71 with 2021-07-01 is defensible — but it is not an AI-specific dated obligation and I have not recorded it as one.

Watching: Finalisation of the National AI Policy in FY2026-27 and any implementing instrument with a commencement date; or an Information Regulator guidance note or enforcement notice on AI / automated decision-making carrying a stated compliance date.

《数字虚拟人信息服务管理办法（征求意见稿）》 (draft Administrative Measures for Digital Virtual Human Information Services) remains AT DRAFT/CONSULTATION STAGE ONLY as of 2026-08-30. The CAC consultation notice was published 2026-04-03 and states verbatim '意见反馈截止时间为2026年5月6日' (comment deadline 6 May 2026) — so the consultation did close on 6 May 2026 as reported. But NO promulgated version has been issued since. Checked on 2026-08-30: (1) the draft's own effective-date article is LITERALLY BLANK — Art. 27 reads '本办法自2026年 月 日起施行。' with the month and day left as unfilled placeholders, so there is no date to record even in the draft; (2) a site-restricted search of cac.gov.cn for 数字虚拟人 returns only the 2026-04-03 consultation-stage documents (the notice plus two 专家解读 expert-commentary pieces) and nothing later; (3) no CAC order (令) promulgating these measures was found for June, July or August 2026. Contrast with the sibling instrument 《人工智能拟人化互动服务管理暂行办法》, which DID complete the cycle — consultation 2025-12-07, promulgated as Order No. 21 on 2026-04-10, effective 2026-07-15 — showing what promulgation of one of these looks like when it happens.

Watching: Publication on cac.gov.cn of the final 《数字虚拟人信息服务管理办法》 as a numbered CAC order (令), which will carry a filled-in Art. 27 in the form 本办法自YYYY年M月D日起施行. Take the date from that article verbatim — do NOT infer it from the consultation close, and do NOT assume the ~3-month promulgation-to-effect gap seen in the anthropomorphic-interaction measures. Watch also for whether it is issued by CAC alone or jointly with other ministries, since that changes who enforces it.

As at 2026-08-30 Indonesia has NO promulgated AI-specific instrument and therefore no dated AI duty. What I checked: (1) The two AI Presidential Regulations — RPerpres on AI Ethics (Etika Kecerdasan Artifisial) and RPerpres on the National AI Roadmap — were listed in Presidential Decision (Keppres) No. 38 of 2025 of 22 December 2025 setting the 2026 Perpres drafting programme, but as of mid-August 2026 both are still UNSIGNED and awaiting the President; Komdigi's own JDIH item of 12 February 2026 says only 'Pemerintah menargetkan Rancangan Perpres Peta Jalan Kecerdasan Artifisial Nasional dapat ditetapkan pada kuartal pertama tahun 2026' ('The Government targets that the draft Presidential Regulation on the National AI Roadmap can be enacted in Q1 2026') — a target that slipped, and a target is not a promulgated date. (2) PP No. 33 of 2026, the long-awaited implementing regulation of the Personal Data Protection Law (UU 27/2022), was enacted and promulgated on 16 July 2026 (Lembaran Negara 2026 No. 88, Tambahan LN No. 7190; 12 chapters, 225 articles) — but it comes into force '6 months after promulgation', a PERIOD with no calendar date stated in the instrument or by Komdigi, so I did not derive 16 January 2027; and I could not confirm that its operative text names AI or solely-automated decision-making (the AI hook sits in Art. 10 of the parent UU PDP, on objecting to decisions based solely on automated processing including profiling). (3) PP No. 17 of 2025 (PP TUNAS, governance of electronic systems in child protection) was enacted and promulgated 27 March 2025 with a transition expressed as a period, and is not AI-specific. (4) Komdigi Circular Letter (Surat Edaran) No. 9 of 2023 on AI ethics is guidance and creates no enforceable duty.

Watching: Presidential signature on either AI Perpres (Etika KA; Peta Jalan KA Nasional) and its promulgation date in the Lembaran Negara; a Komdigi Ministerial Regulation implementing them; the announced move from Perpres to a full UU AI; and any Komdigi or OJK statement fixing the calendar date on which PP 33/2026 becomes effective.

Japan's AI Promotion Act (人工知能関連技術の研究開発及び活用の推進に関する法律, Act No. 53 of 4 June 2025) is fully in force but is PROMOTIONAL, not regulatory — it creates no dated compliance duty and carries no sanction, so it does not qualify as a calendar row. Verified against the official e-Gov statute text (API, law id 507AC0000000053): 附則第一条 reads 'この法律は、公布の日から施行する。ただし、第三章及び第四章並びに附則第三条及び第四条の規定は、公布の日から起算して三月を超えない範囲内において政令で定める日から施行する。' ('This Act comes into effect on the date of promulgation; provided that Chapters 3 and 4 and Arts. 3 and 4 of the Supplementary Provisions come into effect on a date specified by Cabinet Order within a period not exceeding three months from the date of promulgation.'). The Cabinet Office states: '令和7年6月4日にＡＩ法が公布・一部施行され、９月１日にはＡＩ戦略本部の設置に係る規定等も含め、全面施行されました' ('The AI Act was promulgated and partly brought into force on 4 June 2025, and on 1 September it was fully brought into force, including the provisions establishing the AI Strategy Headquarters'). The e-Gov text contains NO article headed 罰則 (penal provisions). The only business-facing article, Art. 7 活用事業者の責務 (duties of AI-utilising business operators), is a best-efforts duty to pursue AI use and to cooperate with state and local government measures — unenforceable by sanction. The AI Basic Plan (人工知能基本計画 ～「信頼できる AI」による「日本再起」～) was adopted by Cabinet decision on 23 December 2025 (cited in the PPC's own 2026 APPI overview) and is likewise a plan, not an obligation. As at 2026-08-30 Japan still has no statutory duty to label AI-generated content, and no dedicated deepfake statute — the government said it would assess sexual-deepfake measures during FY2026 under the AI Promotion Act.

Watching: Any amendment adding sanctions to the AI Promotion Act; a dedicated deepfake/synthetic-media bill in the Diet; a Cabinet Order fixing the enforcement date of the 2026 APPI amendment (separate watchlist row); FSA or PPC binding notices naming AI.

Malaysia's cross-sector AI Governance Bill (Rang Undang-Undang Tadbir Urus Kecerdasan Buatan) is NOT enacted and carries no date. The Ministry of Digital (Kementerian Digital) opened public engagement on the proposal via the Unified Public Consultation portal on 10 July 2026, following the Digital Minister's remarks in Parliament on 24 June 2026; the press release states no date for tabling in the Dewan Rakyat and no commencement date. It would be Malaysia's first horizontal, risk-based AI legal framework with safeguards and incident reporting. Separately, the National Guidelines on AI Governance & Ethics and the National AI Office (NAIO) framework are non-binding. The one binding, dated AI-relevant Malaysian duty found is the MCMC Risk Mitigation Code under the Online Safety Act 2025, recorded as a deadline row (enforcement 1 June 2026, synthetic-media labelling at para. 4.2.4(d)). The Online Safety Act 2025 (Act 866) itself came into operation on 1 January 2026 but its operative text does not name AI — its AI reach comes through the Code.

Watching: Tabling and passage of the AI Governance Bill in the Dewan Rakyat and its commencement clause; any further MCMC code or determination under the Online Safety Act 2025 expressly requiring AI-content labelling (MCMC has said it is considering making AI-content labels mandatory); and subsidiary legislation under Act 866 such as the Online Safety (Online Safety Plan) Regulations 2026.

Beyond the Biometric Processing Privacy Code 2025 (recorded separately as dated rows), New Zealand has NO binding AI instrument and no dated AI duty. The Algorithm Charter for Aotearoa New Zealand (Stats NZ) is an explicitly voluntary commitment signed by government agencies about their own algorithm use — not law, and it binds no private party. The 2025 National AI Strategy and the public-sector generative-AI and business 'responsible AI' guidance are non-binding. New Zealand's stated policy is that existing technology-neutral law (principally the Privacy Act 2020) is sufficient, and no omnibus AI bill has been introduced.

Watching: Any AI bill introduced to Parliament; a further Privacy Commissioner code of practice or compliance notice aimed at automated decision-making; Financial Markets Authority or Reserve Bank instruments naming AI with a compliance date.

As at 2026-08-30 there is NO signed Philippine AI statute. The House is consolidating 26 bills, 3 resolutions and a privilege speech into a draft Artificial Intelligence Development and Regulation Act (building on HB 7396 Artificial Intelligence Development Act and HB 7913 AI Bill of Rights, which would create an AI Development Authority); all remain under legislative review. The National AI Strategy for the Philippines (NAIS-PH), approved by the President in May 2025, is a strategy, not an obligation. The one genuinely binding, dated, AI-specific Philippine instrument I found is COMELEC Resolution No. 11064 (guidelines on the use of social media, artificial intelligence and internet technology for the 2025 National, Local and BARMM Parliamentary Elections), promulgated 17 September 2024, effective 26 September 2024, amended by Resolution No. 11064-A of 13 November 2024. It banned 'false amplifiers' (fake accounts, bots, deepfakes, cheapfakes, softfakes), required disclosure of AI use in campaign material and required registration of digital campaign platforms with the COMELEC Education and Information Department within 30 calendar days of filing certificates of candidacy, i.e. by 13 December 2024. I have NOT recorded it as a live calendar row because it is expressly tied to the May 2025 election cycle, which has passed, and no successor resolution has yet been issued for the 2028 cycle. Sanctions were criminal charges under the Omnibus Election Code, including disqualification.

Watching: Passage and signing of the consolidated Artificial Intelligence Development and Regulation Act (a Republic Act would state its own effectivity clause); a fresh COMELEC resolution on AI for the 2028 elections; and any National Privacy Commission circular (as opposed to advisory) imposing dated AI or automated-decision duties under the Data Privacy Act.

Apart from the two dated rows recorded (Elections (Integrity of Online Advertising) (Amendment) Act 2024 in force 22 January 2025, and the partial commencement of the Online Safety (Relief and Accountability) Act 2025 on 29 June 2026), Singapore has NO binding, dated AI instrument. Specifically: (1) The Model AI Governance Framework — including the Generative AI edition — is VOLUNTARY, an IMDA/AI Verify Foundation advisory framework with no legal force and no compliance date; it is not a calendar row. (2) MAS's Guidelines on AI Risk Management are supervisory expectations, not law: MAS consulted on them from 13 November 2025 with comments due 31 January 2026, proposed a 12-month transition after issue, and as at August 2026 the final Guidelines had not been issued — so there is no date, and even when issued they are Guidelines, not a Notice (MAS Notices are the legally binding form). (3) The Government has expressly declined to impose AI labelling: MTI's written reply to a Parliamentary Question of 6 November 2025 states 'There are currently no plans to introduce specific disclosure requirements or labelling standards for AI-generated content of products and services', relying instead on Technical Reference 76, the AI Markets Toolkit and advertising standards — all voluntary. (4) Health-sector AI guidance from HSA (software/AI medical devices) is regulatory guidance under the existing Health Products Act device-registration regime, with no AI-specific dated duty. (5) Under OSRAA, the deepfake-specific harm 'inauthentic material abuse' (Clause 16) is defined by reference to material 'altered or generated using digital means' and expressly names generative AI, but it is in the deferred tranche with NO commencement date announced.

Watching: The commencement notification bringing OSRAA's 'inauthentic material abuse' and the other seven deferred harms into force; MAS issuing the final AI Risk Management Guidelines (which would start their 12-month transition) or converting any AI requirement into a binding Notice; any IMDA Code of Practice amendment under Part 10A of the Broadcasting Act adding AI-content duties; any reversal of MTI's 'no plans' position on AI labelling.

Beyond the AI Basic Act itself (recorded as a dated row: promulgated and in force 14 January 2026), Taiwan has NO further binding AI instrument carrying a date. What I checked: (1) The Act contains no penal provisions; its Art. 18 obliges GOVERNMENT to complete the adaptation of laws and administrative measures 'within two years of this Act coming into force' — a period, not a stated date, and a duty on the state, not on business, so I did not derive 14 January 2028. (2) The MODA (數位發展部) AI risk-classification framework required by Art. 16 — the instrument that would actually put concrete duties on providers of high-risk AI — has been announced but not yet published, and no publication or compliance date has been stated. (3) The FSC's '金融業運用人工智慧（AI）指引' (Guidelines for Financial Institutions' Application of AI) are guidelines, not a binding order (法規命令); the FSC Chairman said in May 2026 that programmable AI, AI agents and AI risk classification would be folded into those Guidelines, again with no stated compliance date. (4) The legislative supplementary resolutions attached to the AI Basic Act set 3-month, 6-month and 1-year tasks (gender/human-rights/child impact assessments by MODA with NSTC, MOHW, MOE and the Executive Yuan human-rights and gender-equality offices; agency AI risk assessments and internal control rules; MOE AI-use and learning guidance) — all periods, all directed at government bodies.

Watching: MODA's publication of the AI risk-classification framework and any sectoral management rules (管理規範) made under Art. 16 — those would carry the first dated duties on private AI providers; conversion of the FSC AI Guidelines into a binding order; and the government's law-adaptation package under Art. 18.

As at 2026-08-30 Thailand has NO enacted AI instrument and therefore no dated AI duty. The Ministry of Digital Economy and Society (DE) and ETDA published '(ร่าง) หลักการของกฎหมายว่าด้วยปัญญาประดิษฐ์' — DRAFT principles for Thailand's first AI law, focused on controlling high-risk AI — and ran a public hearing that closed 24 June 2568 (2025). ETDA's own page confirms it is still a draft (ร่าง): it has NOT been published in the Royal Gazette (ราชกิจจานุเบกษา) and states no enactment or entry-into-force date. A companion draft Royal Decree on AI-system business services completed consultation but likewise has not been gazetted. Reporting indicates the government aims to move the draft through Cabinet and Parliament during 2569 (2026), but a target is not a promulgated date. The National Press Council's 2567 (2024) guidance on AI and journalistic ethics is a professional-ethics guideline, not law.

Watching: Publication of the AI law or the Royal Decree on AI business services in the Royal Gazette (ราชกิจจานุเบกษา) — Thai instruments state their entry into force relative to that gazette date; also watch PDPC subordinate notifications and BOT/SEC notifications naming AI.

The EU AI Act has not been taken into the EEA Agreement, so Iceland has no implementing act and no date. Verified on the Icelandic Government's own site (Stjornarradid, joint release of the Ministry of Culture, Innovation and Higher Education, the Ministry of Justice and the Ministry of Industries, dated 28 May 2026), which says of the AI Act: 'Gert er rad fyrir ad leggja fram frumvarp a Althingi til innleidingar a reglugerdinni thegar hun hefur verid tekin upp i EES-samninginn' - a bill will be laid before Althingi to implement the regulation ONCE it has been taken into the EEA Agreement. The same release confirms the DSA likewise 'hefur ekki verid tekin upp i EES-samninginn'. No Icelandic AI bill has been introduced and no commencement date exists.

Watching: An EEA Joint Committee Decision incorporating Regulation (EU) 2024/1689 into the EEA Agreement, followed by an Icelandic implementing bill (frumvarp) with a stated gildistaka (entry into force). Also watch the three-ministry deepfake working group set up on 28 May 2026, whose conclusions are due 'eigi sidar en 1. oktober' (no year stated in the source, so not recorded as a dated row) - its recommendations could produce dated Icelandic deepfake legislation.

The EU AI Act has not been incorporated into the EEA Agreement, so it is not applicable in Liechtenstein and there is no national implementing act or date. Primary evidence opened live: the Government's Interpellationsbeantwortung to the Landtag on artificial intelligence, cloud computing and technological infrastructure, BuA No. 59/2026 (Vaduz, 12 May 2026, LNR 2026-660, on the Government's own Berichte und Antraege database), where the interpellating MPs ask the Government: 'Welchen Zeitplan sieht die Regierung fuer die EWR-Uebernahme relevanter EU-Rechtsakte (z. B. EU-KI-Verordnung/AI Act, NIS2), und welche Auswirkungen erwartet sie fuer Unternehmen und Verwaltung?' - i.e. as of May 2026 no timetable for EEA incorporation had been published, which is why Parliament had to ask. Corroborated by Norway's Nkom (page updated 24.03.2026: the regulation must be taken into the EEA Agreement before it has effect) and by the Icelandic Government (28 May 2026: a bill will be laid 'once it has been taken up into the EEA Agreement'). Separately, in April 2026 the Government adopted only an internal AI strategy for the national administration (Medienmitteilung 21.04.2026), which is a policy document with no fixed validity period and no obligations on private parties; the only date it contains is a review 'spaetestens im Jahr 2030'.

Watching: An EEA Joint Committee Decision incorporating Regulation (EU) 2024/1689 into the EEA Agreement, and the Government's subsequent Vernehmlassung and Bericht und Antrag for a Liechtenstein implementing act (Durchfuehrungsgesetz) - the BuA will state the intended Inkrafttreten. The Stabsstelle fuer Digitale Innovation (SDI) leads the EEA incorporation and national implementation.

The EU AI Act (Regulation (EU) 2024/1689) has NOT been incorporated into the EEA Agreement, so it does not apply in Norway and there is no EEA-derived date. Verified on the page of Nkom, the Norwegian Communications Authority designated as Norway's AI coordination authority (page last updated 24.03.2026), which states: 'For at dette regelverket skal fa virkning i Norge, ma det tas inn i EOS-avtalen gjennom den nye norske loven om kunstig intelligens (KI-loven).' Nkom's own timeline page lists only EU application dates (1 Aug 2024 entry into force, 2 Feb 2025 prohibitions, 2 Aug 2025 GPAI, 2 Aug 2026 remaining provisions) and makes no statement that any of them apply in Norway. Nkom also records that, following the AI omnibus, the date for the high-risk requirements is 'not determined'. Nationally: the draft Norwegian AI law (KI-loven) went to consultation with a 30 September 2025 deadline; after the EU simplification package the government said it would run a second consultation and its stated ambition is to lay the bill before the Storting in spring 2027 - an ambition, not an appointed date. I could not open efta.int directly (Cloudflare human-verification challenge, which I did not bypass); its EEA-Lex factsheet for 32024R1689 is reported as 'under scrutiny for incorporation into the EEA Agreement by Iceland, Liechtenstein and Norway', consistent with the regulator's statement.

Watching: Adoption of an EEA Joint Committee Decision incorporating Regulation (EU) 2024/1689 (and Regulation (EU) 2026/1744, the AI omnibus) into Annex XI of the EEA Agreement - the JCD itself states its entry-into-force date and any adapted transitional dates, which is what would create the first real Norwegian calendar rows. Also watch: Royal assent and a commencement date for the Norwegian KI-lov, and the opening of the announced second consultation (which will carry a stated closing date).

Checked the Federal Council's own media release of 12 February 2025 (admin.ch, opened directly) - it sets only an internal planning milestone: 'Das EJPD wird mit dem UVEK und dem EDA bis Ende 2026 eine Vernehmlassungsvorlage erstellen' and 'Das UVEK wird zudem mit dem EJPD, dem EDA und dem WBF bis Ende 2026 einen Plan fuer die weiteren Massnahmen von rechtlich nicht verbindlicher Natur erarbeiten'. 'By end of 2026' is a departmental deliverable, not a stated date and not a duty on any regulated entity; the approach is expressly sectoral with no horizontal Swiss AI act. On the treaty side: the Council of Europe Treaty Office chart for CETS No. 225 (Framework Convention on AI), 'Status as of 30/08/2026', shows Switzerland signed 27/03/2025 but has NOT ratified; 'Total number of ratifications/accessions 1' (European Union, 15/05/2026) against an entry-into-force rule of '5 Ratifications including at least 3 member States of the Council of Europe' - so the Convention is not in force and creates no dated Swiss obligation. Also checked admin.ch/news for any 2026 Federal Council AI decision superseding the 12 Feb 2025 one - none found. Swiss sectoral automated-driving ordinance (VAF) has been in force since 1 March 2025, i.e. not a recent or forthcoming milestone.

Watching: Publication of the EJPD/UVEK/EDA consultation draft (Vernehmlassungsvorlage) implementing the CoE AI Convention - the opening of a Vernehmlassung carries a stated closing date, which would be the first real Swiss AI calendar date; a Federal Council dispatch (Botschaft) to Parliament on ratifying CETS No. 225; deposit of the 5th ratification of CETS No. 225 (3 of them CoE members), which would fix the Convention's entry-into-force date.

Companion entry to the six dated UK rows in this file - these AI items were checked and have NO stated date, so they are deliberately NOT calendar rows. (1) Regulating for Growth Bill: announced in the King's Speech of 13/14 May 2026, creating cross-economy regulatory sandbox powers including for AI; not yet introduced with a published text and no commencement dates - the 2026 legislative programme was set out in the written statement of 14 May 2026 (HCWS7). (2) ICO statutory AI/ADM code of practice: SI 2026/425 imposes the duty to prepare it from 12 May 2026 but sets NO deadline; the ICO says timings will come via its codes of practice pipeline and draft ADM/profiling guidance will be consulted on first. (3) Crime and Policing Act 2026 s.72 (child sexual abuse image-generators, new SOA 2003 s.46A - 'thing' includes 'a program, information in electronic form and a service'): commencement note still reads 'not in force at Royal Assent, see s. 255(1)' and it was absent from the Commencement No.1 list; no appointed day. (4) Automated Vehicles Act 2024 Part 1 (the core self-driving vehicle authorisation regime) plus s.84 civil sanctions and s.89(8)(b)/(10): uncommenced, no appointed day; DfT has publicly referred only to 'the second half of 2027', which is not a stated date. (5) DUAA 2025 ss.135-136 (economic impact assessment on copyright and AI, and report on the use of copyright works in the development of AI systems): the duties are expressed as 'before the end of the period of 9 months beginning with the day on which this Act is passed' - a period, not a stated date, so no row is created. (6) FCA: confirmed on its own 'AI and the FCA: our approach' pages that it is not introducing AI-specific rules, relying on the Consumer Duty and SM&CR; its 2026 AI activity was input-gathering (long-term review contributions closed 24 February 2026; AI Input Zone open 14 May to 19 June 2026), not compliance duties. (7) MHRA: software and AI as a medical device reform remains at consultation/evidence stage with no AI-specific SI in force or with an appointed date. (8) Ofcom: has published guidance that AI chatbot output falls inside Part 3 duties only where it is user-generated or search content - interpretation of existing 2025 deadlines, not a new dated AI duty. (9) Council of Europe Framework Convention on AI (CETS No. 225): the UK signed on 05/09/2024 but has not ratified; per the CoE Treaty Office chart, status as of 30/08/2026, there is 1 ratification in total and the treaty is not in force.

Watching: Introduction and commencement provisions of the Regulating for Growth Bill; an ICO announcement of the AI/ADM code timetable (draft for consultation carries a stated closing date); a Crime and Policing Act 2026 commencement SI appointing a day for s.72; an Automated Vehicles Act 2024 commencement SI for Part 1; and above all a draft statutory instrument under the new s.216A of the Online Safety Act 2023 bringing AI services and illegal AI-generated content into scope - laying that draft before 31 December 2026 would displace the s.249 progress-report duty already recorded as a dated row.

Checked the Council of Europe Treaty Office chart of signatures and ratifications for CETS No. 225 directly on 2026-08-30; the page itself is stamped 'Status as of 30/08/2026'. It reports 'Total number of ratifications/accessions 1' and 'Total number of signatures not followed by ratifications 20'. The ONLY ratification is by the European Union (signature 05/09/2024, ratification 15/05/2026) — an international organisation, not a State. ZERO Council of Europe member States have ratified. The Entry into Force condition printed in the chart header is '- 5 Ratifications including at least 3 member States of the Council of Europe', and Article 30(3) of the Convention (as published in OJ L 2026/1081, 13.5.2026) requires five signatories including at least three CoE member States. The 'Entry into Force' column is BLANK for every row, including the EU's. Therefore the Convention has NOT entered into force, it binds nobody yet, and no entry-into-force date exists to record — not for the treaty and not for the EU. NOTE FOR THE EDITOR: widely repeated secondary commentary claiming CETS 225 'entered into force on 1 November 2025 after ratification by the United Kingdom, France and Norway' is FALSE against the depositary chart — the UK and Norway show signature only (both 05/09/2024) and France appears in the chart with no signature and no ratification at all.

Watching: Four more ratifications, at least three of which must be by Council of Europe member States. When the fifth consent-to-be-bound is deposited, Article 30(3) fixes general entry into force on 'the first day of the month following the expiration of a period of three months' after that deposit; each later ratifying State then gets its own staggered date under Article 30(4). Re-check the same Treaty Office chart — it publishes an 'Entry into Force' column per State and the Secretary General notifies EIF dates under Article 36(c). Do NOT compute any of these dates: take them from the chart once printed.

NO ratification anywhere creates a dated domestic implementation duty at present, and the reason is twofold. FIRST: no STATE has ratified CETS No. 225 at all — the Council of Europe Treaty Office chart, status as of 30/08/2026, reports 'Total number of ratifications/accessions 1', and that single ratification is the European Union (signed 05/09/2024, ratified 15/05/2026). So there is no national ratification anywhere from which a domestic implementation duty could flow. SECOND: the EU's own ratification instrument expressly declines to create one. Council Decision (EU) 2026/1080 of 21 April 2026 on the conclusion, on behalf of the European Union, of the Convention states at Article 3: 'The Convention shall be implemented in the Union exclusively through Regulation (EU) 2024/1689 and other relevant Union acquis, where applicable.' That is a deferral, not a new duty — the word 'exclusively' rules out any separate implementation obligation or timetable, and Article 4 provides only that 'This Decision shall enter into force on the date of its adoption.' Nothing in the Decision imposes a dated obligation on member states or on companies; every operative date for EU-regulated entities remains an AI Act date and is already tracked under that instrument. Compounding all of this, the Convention is not in force (it needs five ratifications including at least three Council of Europe member States), so even the EU is not yet bound by it and its 'Entry into Force' cell in the depositary chart is blank. Conclusion: NO calendar row is justified for any ratification at this time.

Watching: Two distinct triggers, and they are not the same event. (1) The Convention's general entry into force, once the fifth consent-to-be-bound is deposited with at least three of them from CoE member States — read the date off the Treaty Office chart's 'Entry into Force' column, never compute it. (2) A ratifying STATE that, unlike the EU, enacts implementing legislation with its own commencement date — that domestic statute, not the ratification, would be the calendar row. Watch in particular any ratifying state without an existing horizontal AI statute, since those are the ones likely to legislate afresh. For the EU specifically, treat the AI Act timetable as the only live source of dates; do not create CETS 225 rows for EU operators.

NOT BINDING — verified against UNESCO's own Legal Affairs page for the instrument. The UNESCO Recommendation on the Ethics of Artificial Intelligence was adopted on 23 November 2021 at the 41st session of the General Conference in Paris. A 'Recommendation' is UNESCO's non-binding class of standard-setting instrument, as distinct from a Convention, and the text itself makes the voluntary character explicit: Member States are to 'apply on a voluntary basis the provisions of this Recommendation by taking appropriate steps, including whatever legislative or other measures may be required, in conformity with the constitutional practice and governing structures of each State'. There is NO dated compliance deadline for States and none at all for companies — the Recommendation does not bind private actors in any way. Implementation tooling (the Readiness Assessment Methodology, the Ethical Impact Assessment) is advisory and carries no deadline. UNESCO's role is described as partnership and support: 'UNESCO can be a partner and support Member States in the development as well as monitoring and evaluation of policy mechanisms.' Nothing here is a calendar row.

Watching: Only a binding successor instrument would qualify — i.e. UNESCO's General Conference adopting a CONVENTION on AI (which would then need signature, ratification and an entry-into-force clause), rather than revising or supplementing the Recommendation. Note that even the periodic implementation reporting UNESCO Member States do under the Constitution is a reporting cycle to UNESCO, not a compliance duty on regulated entities, so it should not be converted into calendar rows.

NOT BINDING — read the primary text of UN General Assembly resolution A/RES/79/325, 'Terms of reference and modalities for the establishment and functioning of the Independent International Scientific Panel on Artificial Intelligence and the Global Dialogue on Artificial Intelligence Governance', headed 'Resolution adopted by the General Assembly on 26 August 2025' (89th plenary meeting, 26 August 2025). A UNGA resolution of this kind is a recommendation, and the text confirms it on its face: the operative verbs create UN INSTITUTIONS, not duties on anyone — para 1 'Establishes, within the United Nations, the multidisciplinary Independent International Scientific Panel on Artificial Intelligence', para 4 'Establishes, within the United Nations, the Global Dialogue on Artificial Intelligence Governance'. The Panel's output is expressly described as 'one annual policy-relevant but non-prescriptive summary report'. The ONLY operative paragraph aimed at States is para 11, which merely 'Encourages States, as well as the private sector, financial institutions, foundations and other donors in a position to do so, to support the effective functioning of the Panel and Dialogue'. 'Requires' (para 2) binds candidates for Panel nomination to disclose conflicts of interest, not Member States; 'Requests' (paras 3, 10) is addressed to the Secretary-General. There is NO obligation on any State or company and NO dated compliance deadline anywhere in the resolution. Its predecessor instruments are the same in character: resolution 79/1 'The Pact for the Future' and its annex the 'Global Digital Compact', plus resolutions 78/265 and 78/311, all recommendatory. The dates the resolution does contain are EVENT dates for UN meetings, not compliance dates: para 6 provides the Global Dialogue 'will initially be held back-to-back in the margins of the International Telecommunication Union Artificial Intelligence for Good Global Summit in Geneva, in 2026, and of the multi-stakeholder forum on science, technology and innovation for the Sustainable Development Goals in New York, in 2027'. Related 2026 activity is also institutional, not regulatory: the GA appointed the Panel's 40 members on 12 February 2026 and the first Global Dialogue was held in Geneva on 6-7 July 2026. None of this belongs in a compliance calendar.

Watching: The UN has no treaty-making step under way for AI. A calendar row would require a genuinely binding instrument — a convention opened for signature and ratification under a UN depositary, with an entry-into-force clause — not another resolution, dialogue or panel report. Para 12 notes only that continuation of the Panel's and Dialogue's terms of reference 'may be considered and decided upon by the General Assembly during the high-level review of the Global Digital Compact at its eighty-second session'; that is a mandate-renewal question, still not an obligation on anyone.

Argentina has NO AI statute. Checked 2026-08-30. The closest binding-adjacent instrument is Resolucion SIGEN 197/2026 (Sindicatura General de la Nacion, Argentina's internal-audit body), dated 16 June 2026 and published in the Boletin Oficial No. 35934 on 22 June 2026, page 29, whose sole operative clause is 'APRUEBASE LA GUIA DE CONTROLES INTELIGENCIA ARTIFICIAL - SIGEN'. It is a risk-based internal-control GUIDE for AI adoption across the National Public Sector, not a regulation imposing duties on regulated entities, and — checked on the official Argentina.gob.ar normativa record — it contains NO vigencia article and NO compliance deadline. Several AI bills are in parliamentary discussion, informed by an anteproyecto prepared by the Agencia de Acceso a la Informacion Publica (AAIP), but none has been sanctioned. In the meantime AI that processes personal data falls under the pre-existing Ley 25.326 de Proteccion de los Datos Personales, which dates from 2000 and gained no new date from any of this.

Watching: Sanction and Boletin Oficial publication of an AI law, or an AAIP resolution (rather than an anteproyecto) imposing dated duties on automated decision-making. A SIGEN follow-up converting the guide into mandatory audit criteria with a compliance date would also create a row.

ANPD (Autoridade Nacional de Protecao de Dados) has issued NO binding normative act on AI carrying a compliance date. Checked 2026-08-30. What exists is agenda and preparatory work only: item 7 of the ANPD Regulatory Agenda for the 2025-2026 biennium is dedicated to AI and specifically to the right to review automated decisions under LGPD art. 20; the ANPD ran a Tomada de Subsidios on that topic and consolidated the responses in Nota Tecnica n. 12/2025; and in December 2025 the ANPD published its Priority Themes Map for 2026-2027 naming artificial intelligence and emerging technologies as one of four supervisory axes. None of these is a regulation and none carries an obligation date. The underlying LGPD duty — art. 20, the data subject's right to request review of decisions taken solely on automated processing of personal data affecting their interests, and the controller's duty to give clear information about the criteria and procedures used — has been in force since the LGPD took effect in 2020 and was NOT given a new date by any of this work.

Watching: Publication in the Diario Oficial da Uniao of an ANPD resolution regulating LGPD art. 20 / automated decisions, which would carry its own vigencia clause and any transition period. Also watch whether PL 2338/2023 designates ANPD as the coordinating authority of the SIA, which would trigger implementing regulations with dates.

Chile's dedicated AI bill — 'Regula los sistemas de inteligencia artificial', Boletin 16821-19 (refundido with 15869-19), a risk-based framework imposing obligations on developers, providers, implementers and distributors of AI systems including foreign ones operating in Chile — is NOT law. Verified 2026-08-30: the Camara de Diputados approved it en particular on 13 October 2025 and remitted it to the Senate; it is in SECOND constitutional trámite before the Senate's Comision de Desafios del Futuro, Ciencia, Tecnologia e Innovacion, and must also pass the Comision de Hacienda. Recorded movements include an urgencia suma presented 6 January 2026 and withdrawn-and-re-presented 20 January 2026. No third trámite, no approval, no promulgation, no Diario Oficial publication — therefore no commencement date exists. Note this is SEPARATE from Ley 21.719 (data protection), which IS enacted and does have a dated commencement — recorded as a deadline row.

Watching: Senate approval in second trámite, any third trámite or comision mixta, then promulgation and Diario Oficial publication with the transitional article setting vigencia. The BCN record at bcn.cl will show 'Entra en vigencia el ...' once published, which is the citable date source.

Colombia has NO enacted AI law and no binding regulator instrument on AI with a date. Checked 2026-08-30. The government-backed bill PL 043/2025 Senado - 324/2025 Camara ('Por medio del cual se regula la inteligencia artificial en Colombia...'), filed 28 July 2025 by MinCiencias and given an urgency message on 8 September 2025, was ARCHIVED at the end of the legislature without becoming law. A replacement bill was radicado on 21 July 2026 in the Camara de Representantes — 'Proyecto de Ley Por medio de la cual se regula la inteligencia artificial en Colombia para garantizar su desarrollo etico y responsable y se dictan otras disposiciones' (proyecto 36127, P.L. 025-2026SC), an 86-page text. Its own Article 37 reads 'Entrada en vigencia. La presente ley regira a partir de su promulgacion y deroga...' — commencement on promulgation, so even the bill states no date, and it has not been debated, approved or promulgated. Colombia's AI policy instrument, CONPES 4144, is a policy document setting government action lines, not a source of enforceable dated duties on regulated entities.

Watching: Passage of P.L. 025-2026SC through both chambers, presidential sanction and publication in the Diario Oficial — at which point Article 37's 'a partir de su promulgacion' plus the promulgation date would give a citable date. Also watch the Superintendencia de Industria y Comercio (data protection authority) for any circular externa on automated decision-making, which would carry its own effective date.

There is NO Ley General de Inteligencia Artificial in Mexico. Checked 2026-08-30: the comprehensive AI bill never reached the floor of Congress, so it has not been voted and has not been published in the Diario Oficial de la Federacion. Separately, a constitutional reform initiative was presented to the Comision Permanente by Morena deputy Gabriela Jimenez to amend fraccion XVII of Article 73 of the Constitution so as to expressly empower the Congreso de la Union to legislate on the use and implementation of artificial intelligence systems — that is an enabling-power amendment, still an initiative, not approved and not published. Neither carries any date. Mexico's AI regulation is therefore sectoral only (see the separate entry on the LFT/LFDA reform of 14 May 2026).

Watching: Approval and DOF publication of the Article 73 constitutional reform (which would then require secondary legislation), or a Ley General de IA reaching a floor vote. Either would come with transitorios setting commencement. Track the Camara de Diputados boletines and the Senado's agenda.

Peru HAS a binding, AI-specific regulation with a full staggered compliance schedule — Decreto Supremo N. 115-2025-PCM approving the Reglamento of Ley N. 31814 (Ley que promueve el uso de la inteligencia artificial en favor del desarrollo economico y social del pais). El Peruano states 'Fecha de publicacion: 09/09/2025'. It creates real duties: risk classification with prohibited uses, Art. 25 'Transparencia Algoritmica' (developers and implementers of HIGH-RISK AI systems must give users prior, clear and simple information about the system's purpose, main functionalities and the type of decisions it can make), and Arts. 28.11 / 31.4 requiring human oversight mechanisms able to stop, correct or invalidate AI decisions in health, education, justice, finance and access to basic services. BUT NOT ONE COMPLIANCE DATE IS WRITTEN AS A CALENDAR DATE — every deadline is a relative period. Vigencia clause: 'El presente Decreto Supremo entra en vigencia a partir de los noventa (90) dias habiles siguientes de su publicacion en el diario oficial El Peruano, con excepcion de la Primera, Segunda, Cuarta y Quinta Disposiciones Complementarias Finales ... las cuales entran en vigencia al dia siguiente de la publicacion'. PRIMERA Disposicion Complementaria Final (Cronograma de implementacion) then sets, all 'contado a partir del dia siguiente de la publicacion del Decreto Supremo': PUBLIC SECTOR — Executive/Legislative/Judiciary 1 year; Organismos Constitucionales Autonomos 1 year; EsSalud, regional governments and public universities 2 years; local governments Type A/B/C 3 years; public enterprises of regional/local governments and FONAFE 2 years; other entities under Ley 27444 2 years; local governments Type D-G optional. PRIVATE SECTOR — health, education, justice, security, economy and finance 1 year; transport, commerce and labour 2 years; production, agriculture, energy and mining 3 years; all other uses 4 years. MYPES/innovative startups — small enterprises (sales above 150 UIT up to 1700 UIT) 2 years; microenterprises (up to 150 UIT) 3 years. Commonly cited derived dates (general vigencia 22 January 2026; first private-sector tranche 10 September 2026) are law-firm computations, not text. Per the no-computed-dates rule none of these becomes a calendar row.

Watching: A Resolucion Secretarial from the SGTD (Secretaria de Gobierno y Transformacion Digital) under the SEGUNDA Disposicion Complementaria Final — it is empowered to issue the complementary norms, technical specifications, standards and the Transparencia Algoritmica lineamientos (Art. 25.4), and such a resolution would carry its own publication and effective dates. Also watch for any SGTD implementation calendar that writes the tranche dates out explicitly; that would immediately yield several high-value rows (deadline_type compliance, ai_specific).

CLOSE TO A ROW BUT THE DATE IS COMPUTED, NOT STATED — deliberately withheld. Egypt's Personal Data Protection Law No. 151 of 2020 was finally operationalised by its Executive Regulations (Ministerial Decision No. 81 of 2025). The Regulations DO carry an AI hook: per Baker McKenzie's review of the text, 'The Regulations refer briefly to AI training, with a requirement for processors to handle personal data in accordance with "locally, regionally and internationally recognized principles" when using personal data for AI training and emerging or innovative technologies. Processors must ensure that these technologies are used in a manner that does not cause harm to the data subject.' The compliance clock is real but its end date is nowhere STATED. The PDPL sets a formula, not a date: 'a grace period for compliance extending until one year after the date of the issuance of executive regulations supplementing the law.' The anchor is contested — the Regulations were 'publicly released on 25 December 2025' while 'The formal date of publication is 1 November 2025 (being the date of the official gazette in which the Regulations were circulated).' Baker McKenzie then only ANTICIPATES the result: 'Accordingly, we anticipate that the compliance grace period contemplated in the PDPL will extend until 1 November 2026.' LexAfrica computes 31 October 2026 from the same facts. Two credible firms differ by a day and one expressly hedges — that is proof the date is derived, not declared. No Egyptian regulator (the Personal Data Protection Centre) has published the deadline. Egypt's draft Artificial Intelligence Law remains in the parliamentary process and is not enacted; the National AI Strategy and AI ethics charter are non-binding.

Watching: A statement from Egypt's Personal Data Protection Centre or a ministerial decision fixing the exact end of the PDPL grace period (expected around 1 November 2026) — that would convert this into a dated row with the AI-training provision as its AI hook. Also: enactment of Egypt's draft AI Law.

Checked for binding AI law and for Privacy Protection Authority AI instruments. Israel has NO AI statute and no binding AI instrument with a stated date. (1) 'Currently, there are no specific codified laws, statutory rules or regulations in Israel that directly regulate AI'; as of 2026 there is no Israeli AI Act and companies rely on a patchwork of existing law. (2) The joint Ministry of Innovation, Science and Technology / Ministry of Justice policy on 'Artificial Intelligence Regulation and Ethics' (2023, following a 2022 White Paper) is POLICY GUIDANCE, not binding law, and it deliberately favours soft regulation — sandboxes, pilots, standards and non-binding principles — plus a proposed AI Policy Coordination Centre to advise sectoral regulators. (3) The Privacy Protection Authority has published DRAFT guidelines on how the Privacy Protection Law applies to AI systems following Amendment 13; these were still draft, with finalisation and active enforcement only expected during 2026 — no adopted date. (4) Amendment 13 to the Privacy Protection Law did come into force in August 2025 and materially expanded PPA enforcement powers, but it is a general privacy amendment, not an AI instrument, and its operative text does not name AI.

Watching: The Privacy Protection Authority finalising its AI guidelines with a stated effective or enforcement date; or an Israeli AI bill reaching the Knesset with a commencement provision.

Checked the Qatar Central Bank 'Artificial Intelligence Guideline on Regulating the use of Artificial Intelligence by QCB Licensed Entities'. Qatar News Agency (official state agency) confirms it was issued on 4 September 2024 ('Doha, September 04 (QNA)'). Its obligations are written in mandatory terms for QCB-licensed entities (AI strategy, risk assessment, human oversight, AI register, approval for high-risk systems, customer transparency, disclosure of AI activity to QCB), so it is binding in substance. BUT no primary source — neither QNA nor the QCB itself — states a separate effective date, compliance deadline or transition period; the duty simply attached on issuance. QCB's own site (qcb.gov.qa) could not be fetched (TLS certificate error). Qatar has NO general AI statute: the National AI Strategy (2019) and the Cabinet Decision No. 10 (2021) AI Committee are strategy/institutional documents, not binding duties. Separately, the QICDRC Practice Direction No. 1 of 2026 IS dated and is recorded as a deadline row.

Watching: A QCB circular setting a dated compliance or attestation deadline for the AI Guideline; or a Qatari federal AI statute or Cabinet decision published in the Official Gazette with a stated in-force date.

Checked every SDAIA instrument plus the Global AI Hub Law, SAMA and the PDPL. NO binding AI instrument with a stated date exists. (1) SDAIA AI Ethics Principles and the SDAIA AI Adoption Framework (v1 Sept 2024, v2 May 2025) are policy guidance — CMS's AI regulation scanner states plainly: 'These instruments are non-binding unless linked to other enforceable laws (e.g., PDPL)', and 'Saudi Arabia does not currently have an AI-specific law and has not announced any formal legislative process to enact one.' Secondary sites describing the Adoption Framework as a 'mandatory baseline' are not supported by any primary text; SDAIA's own PDF (sdaia.gov.sa) is behind a web filter and returns an access-denied page to automated fetching. (2) Global AI Hub Law: still a DRAFT. CST published it for public consultation on 14 April 2025, closing 14 May 2025; CMS confirms 'This draft law has not been enacted to date' and 'It remains a draft and is not yet in force.' It has NOT been published in Umm Al-Qura. Note also that despite its name it regulates data embassies and sovereign hosting, not AI systems. Its draft entry-into-force clause ('shall apply after sixty days from the date of its publication in the Official Gazette') is conditional on a publication date that does not yet exist — computing a date from it would be invention. (3) SDAIA draft Responsible AI Policy: consultation ran 3 April 2026 to 3 May 2026 and had not been adopted as binding as of reporting through July 2026. (4) SAMA has no dedicated AI rulebook chapter with a compliance date; AI expectations sit inside existing cybersecurity and data governance frameworks. (5) PDPL (Royal Decree M/19) does contain automated-decision provisions — a data subject right not to be subject to decisions based solely on automated processing, and mandatory DPIAs for automated decision-making under Article 25 of the Implementing Regulations — but the PDPL is a general data protection statute whose operative text does not name AI or ML, and its compliance date (end of the grace period, 14 September 2024) is a general privacy date, not an AI milestone. 2026 was declared the 'Year of Artificial Intelligence' by Cabinet on 10 March 2026 — a designation, not a duty.

Watching: Publication of the Global AI Hub Law in Umm Al-Qura (would start its stated 60-day clock and produce a real date); formal adoption of SDAIA's Responsible AI Policy as a binding instrument rather than a policy; or a SDAIA/SAMA circular imposing a dated AI compliance or registration obligation.

Checked the AI bills before the Grand National Assembly (TBMM), the KVKK's AI outputs, and the Resmî Gazete. NO binding AI instrument with a compliance date. (1) Legislation: at least three AI legislative proposals have been submitted to the TBMM and NONE has been enacted; the most recent was submitted on 23 July 2025 and remains under committee review. There is no adopted Turkish AI law. (2) The Türkiye Artificial Intelligence Action Plan (2026-2030) WAS published in the official gazette — Presidential Circular No. 2026/9, Resmî Gazete No. 33344 of 18 August 2026 — but it is a national action plan assigning coordination tasks to public institutions under the Ministry of Industry and Technology, not a compliance duty on regulated entities, so it is not a calendar row despite having a real gazette date. (3) KVKK (the Personal Data Protection Authority) has issued AI material in 2026 — a guidance document on 'Agentic Artificial Intelligence (Agentic AI)' published 15 April 2026, plus earlier 'AI Chatbots: Privacy and Security Guidelines' and recommendations on protecting personal data in AI. These are EXPLICITLY NOT BINDING: the agentic AI document 'is not a binding regulation under Law No. 6698 on the Protection of Personal Data, but represents the DPA's formal position on how existing data protection obligations apply'. No KVKK Board decision imposing a dated AI obligation was found.

Watching: Adoption of one of the TBMM AI bills and its publication in the Resmî Gazete with an in-force date; or a binding KVKK Board decision (ilke kararı) on AI carrying a compliance date.

Checked for federal AI legislation, TDRA rules and ADGM. NO binding federal AI instrument with a stated date exists. (1) There is no horizontal UAE AI statute: legal risk for AI is handled through Federal Decree-Law No. 45 of 2021 (PDPL) and sectoral rules. (2) On 14 June 2026 the UAE announced the creation of the Federal Authority for Artificial Intelligence and Data, consolidating AI oversight, digital government and data regulation under one body reporting to Cabinet — this CREATES A REGULATOR, it does not impose a dated duty on anyone, so it is not a calendar row. (3) The UAE AI Charter and the 'Deepfake Guide' published by the UAE National Programme for Artificial Intelligence (July 2021) are expressly ADVISORY — the Deepfake Guide is a non-binding awareness framework complementing existing cybercrime and data protection law. TDRA has issued no binding AI rule with a compliance date. (4) ADGM: the Data Protection Regulations 2021 were enacted 14 February 2021 and applied from 14 August 2021 (new entities) and 14 February 2022 (existing entities), and they do contain a right not to be subject to solely automated decision-making/profiling — but this is a GDPR-style general data protection regulation, not an AI instrument, its operative text does not name AI or machine learning, and its dates are general privacy commencement dates already long past. ADGM's official PDFs (assets.adgm.com and the Thomson Reuters mirror) return HTTP 403 to automated fetching, so the absence of AI wording is asserted from the regulation's structure and secondary summaries, not verified line by line. THE ONE REAL UAE ROW IS DIFC REGULATION 10, recorded separately. IMPORTANT: several sites (uaeahead.com, ailawguide.org, theleveragedyears.com, magureinc.com) assert DIFC Regulation 10 'reached full enforcement on 1 January 2026' — this is unsourced and contradicted by the primary text; do not carry it.

Watching: The new Federal Authority for Artificial Intelligence and Data issuing its first binding regulation or Cabinet Decision with an in-force date; a TDRA rule on synthetic media or deepfakes with a compliance date; or an ADGM AI-specific regulation comparable to DIFC Regulation 10.

Checked 2026-08-30. British Columbia has NO AI statute and no AI regulation. The only province-wide AI instrument is the 'Policy on the use of generative AI' published by the Province on digital.gov.bc.ca, which is an internal conduct policy for BC Public Service employees, not legislation: 'This policy applies to all BC Public Service employees.' It uses mandatory language ('employees must log in using their IDIR', 'must not put any confidential information ... into publicly available gen AI tools like ChatGPT', 'must review the outputs to ensure they are factually accurate'), but it states only 'Last updated on May 9, 2025' — no commencement date, no compliance date, no transition period, and no penalty. It binds employees through the Standards of Conduct and Appropriate Use Policy, not through any dated legal obligation on organisations. BC's FOIPPA contains no automated-decision-making clause equivalent to Quebec s. 12.1 or Alberta POPA s. 5(2)(d).

Watching: A BC statute or FOIPPA amendment creating AI/automated decision duties, or a Treasury Board / Chief Information Officer directive with a stated compliance date. BC's independent oversight offices have publicly called for AI legislation, so a bill is the thing to watch.

Enhancing Digital Security and Trust Act, 2024 (Schedule 1 to Bill 194, Strengthening Cyber Security and Building Trust in the Public Sector Act, 2024, S.O. 2024, c. 24) contains real AI duties for public sector entities — s. 5 (inform the public about AI use, develop and implement an accountability framework, manage risks, comply with prescribed requirements, obey prohibited-use rules) and s. 6 (disclose AI use and ensure a human exercises oversight). But EVERY one of those duties is expressed as 'in accordance with the regulations' / 'as may be prescribed', and s. 5(1) only applies to 'such public sector entities as may be prescribed'. Checked 2026-08-30: the only regulations actually made under EDSTA are O. Reg. 51/26 (cyber security — contact point, biennial cyber maturity assessments, 72-hour incident reporting for hospitals, colleges and universities, school boards and children's aid societies) and O. Reg. 52/26 (digital technology affecting individuals under 18 — plain-language notice by school boards when student digital information is shared with third-party software). The Government of Ontario's own EDSTA page states these take effect 1 July 2026 and neither imposes any AI obligation; AI appears on that page only in the rationale ('address emerging risks from powerful technologies like artificial intelligence') and in a non-statutory internal 'Responsible Use of AI Directive' for ministries and provincial agencies. Bill text s. 17: 'The Act set out in this Schedule comes into force on a day to be named by proclamation of the Lieutenant Governor.' With no AI regulation prescribing entities or circumstances, no AI duty is triggered and no date attaches.

Watching: An Ontario regulation made under EDSTA s. 7 prescribing the public sector entities and circumstances for ss. 5-6, with its own effective date — that is what would create a dated AI obligation. Also watch for the Responsible Use of AI Directive being given statutory force.

Bill C-34 (Safe Social Media Act — enacts the Digital Safety Act and the Digital Safety Commission of Canada Act) would impose a Duty to Protect Children on 'regulated services' expressly including AI chatbot services. LEGISinfo checked 2026-08-30: FIRST READING 2026-06-10, currently 'At second reading in the House of Commons', no further stages completed, NO royal assent. No commencement date stated anywhere in the bill's parliamentary record.

Watching: Royal assent, then the order in council / CIF provision setting the date the Digital Safety Act applies to AI chatbot services, plus the regulations the Digital Safety Commission must make before duties bite.