Canada’s Official Marks Cleanup One Year On, and the Signal from Singapore’s AI Disputes Forum: procedural gateways are tightening and brand plus training-data governance is becoming more verifiable

One year has now passed since the subsection 9(4) mechanism under Canada’s Trademarks Act for official marks came into force on 1 April 2025. For ordinary trade mark applicants, the most important point is not that the official marks regime has suddenly disappeared, but that there is finally a more usable pathway for clearing out legacy official marks that may have long occupied filing space even though the underlying entity may no longer exist or may no longer qualify as a public authority. Upon payment of the prescribed fee and submission of supporting information or evidence, a party may request that the Registrar give public notice that a particular official mark no longer applies; CIPO has also placed the initial step of deciding whether a notice to the holder is necessary within a 12-week service-standard framework.

At almost the same time, discussions linked to IPOS’s participation in I-TIDE 2026 on AI and cross-border technology disputes pushed a different but related frontier issue into view: copyright in AI training data, cross-border dispute design, platform responsibility, and the boundaries of Singapore’s computational data analysis (CDA) exception. Read together, these are not isolated regional updates. They point to a broader institutional direction: one side is procedurally thinning historically over-broad layers of rights blockage, while the other is placing stronger lawful-access, traceability and responsibility demands on the once-comfortable idea that AI actors can train first and explain later. In both brand governance and content governance, what matters increasingly is not merely whether a party can assert a claim, but whether that claim can be checked, evidenced and defended inside an actual procedure.

1. Canada’s official-marks cleanup changes market-entry cost more than it changes the name of the regime

For many businesses running Canadian clearance reviews, one of the most frustrating obstacles has not necessarily been an active prior private trade mark used and enforced in the market, but an older official mark that was once publicly noticed and still blocks later filings even though the holder may have gone quiet for years, changed form, disappeared, or become doubtful as a qualifying public authority. Official marks have long had unusually strong blocking force without fitting neatly into the more familiar cycles of use, renewal and challenge that shape ordinary trade mark practice. The result has been a classic asymmetry: applicants could suspect that an obstacle was effectively dead weight, yet still lack a clear and economically manageable route for testing or removing it.

The real significance of subsection 9(4) is that it begins to convert that asymmetry into a procedural question that can actually be worked with. It does not abolish all official marks, and it does not automatically invalidate older ones. Instead, it creates a more realistic filter: where an official mark is challenged, the logic of continued blockage becomes more reviewable, while the requesting party finally has a defined entry point and some expectation of procedural rhythm. For the market, that matters far more than abstract debates over whether Canada should one day abandon official marks entirely. What businesses care about in practice is whether, after a clearance hit appears, there is a route that allows a project to keep moving.

That said, the Canadian shift should not be misread as simple deregulation. The mechanism still keeps meaningful procedural discipline. A request generally relates to one official mark at a time. The requester must provide concrete information or evidence showing that the holder is not a public authority or no longer exists. If the filing is deficient, the fee is wrong, or the supporting material is missing, the matter can be closed after a short correction window with very limited flexibility. In other words, Canada is not abandoning screening. It is redesigning screening from a largely one-way and opaque barrier into a process gateway that can be triggered, answered and scrutinised.

2. The Singapore AI discussion is not only about whether a model can be used, but on what basis, with what proof, and with whose responsibility

Singapore is often described as AI-friendly, and that description is not baseless. But it becomes misleading if reduced to the idea that AI innovation is simply welcomed on broad permissive terms. The more accurate direction in recent official materials is that Singapore supports data-driven innovation while steadily moving lawful access, technical protection measures, sector-specific accountability and governance transparency closer to the front of the analysis. The topic design of I-TIDE 2026 itself shows that the real centre of gravity is no longer just whether an output may infringe, but also copyright issues in training data, the possibility of secondary responsibility, and how cross-border AI and IP disputes should actually be resolved.

Even more importantly, Singapore’s treatment of the CDA exception is clearly not an unlimited pass. It supports certain non-consumptive uses of copyright-protected material for computational data analysis, including text and data mining and AI model development; but at the same time it places weight on lawful access and rejects the creation of a new exception that would allow circumvention of access-control measures for CDA purposes. That means the policy signal is not “train first, rights holders later.” It is closer to “innovation is permitted within compliance boundaries, but access controls, transaction structures and evidential limits remain real.” As a practical matter, this changes the bargaining landscape between AI developers and content owners. Licensing, controlled-access arrangements, dataset provenance and records of how material was obtained will all become more important.

The broader governance trend in Singapore is also moving beyond copyright exceptions alone. Guidance, testing frameworks and transparency expectations around generative AI are becoming more detailed, especially in higher-responsibility contexts such as legal services, where disclosure, verification, risk classification and human accountability are becoming harder to treat as optional soft issues. For AI service providers, competition is therefore not only about model performance. It is increasingly about documentary readiness, process governance and institutional trust. The party that can show how data was accessed, where system boundaries were set and who remains responsible is more likely to hold a stronger position when a cross-border dispute actually arrives.

3. Read together, Canada and Singapore point to the same turn: broad protective shells are thinning, and procedural verifiability is rising

At first glance, Canada’s story is about official marks, while Singapore’s story is about AI and copyright disputes. They look like different subject areas. But at the level of regulatory method, they are strikingly similar. Canada is narrowing a historically broad, sticky and hard-to-clear blocking space. Singapore is resisting any attempt to package AI training, scraping and deployment as a single black box insulated from differentiated conditions. Both jurisdictions are pushing the market away from a phase dominated by abstract assertions and toward one dominated by procedure, evidence and reviewable accountability.

What does that mean for businesses? It means that the asset of the future is not only a formally broad label of rights or an aggressive external narrative. The more durable advantage lies in whether a business can produce a structured basis for its position when questioned. For parties building brands in Canada, the key question becomes whether an obstructive official mark holder still exists, still qualifies, and still justifies keeping a barrier in place. For businesses deploying AI in or toward Singapore, the question becomes how training data was obtained, whether access was lawful, whether technical controls were bypassed, whether governance records were preserved, and whether dispute clauses have been drafted with cross-border friction in mind.

So this is not merely a convenient pairing of “one trade mark issue and one copyright issue.” It is a deeper rewrite of regulatory grammar. In the past, legal systems were often willing to let broad and static protective shells sit in place for long periods. Now, regulators increasingly expect those shells to survive procedural testing when challenged. In the past, technology narratives often assumed that the market would move first and rules would tidy up later. Now, a more common expectation is that innovation and proof should advance together. For brands, platforms, publishers, database operators and AI providers, that change may matter more than any one statutory amendment because it alters the underlying logic of negotiation, clearance and dispute response.

4. What businesses should prioritise next is not generic policy watching, but a rebuilt evidence map and dispute plan

For brand owners preparing for Canada, the first task is no longer to shrug and treat the existence of official marks as an undifferentiated filing hazard. Clearance results now need to be re-layered: which obstacles are still genuinely active private marks or living official marks, which older official marks deserve further inquiry into holder status and continuing existence, and which may justify a challenge once commercial timing and cost are assessed. In other words, once a cleanup route exists, clearance work is no longer only about spotting an obstacle. It is also about deciding whether the obstacle still deserves to stay in the way. That expands early-stage budget planning beyond search costs into investigation, evidence assembly and procedural strategy.

For businesses using or supplying generative AI services—especially in publishing, media, education, legal tech, enterprise knowledge management and cross-border SaaS—the more valuable next investment is not a high-level policy statement alone. It is the integration of data-source records, lawful-access analysis, licensing arrangements, supplier commitments, customer disclosures, log retention and escalation paths into day-to-day workflow. Many AI disputes may appear to be about outputs at first glance, but in practice they often trace back to the training layer, the way data was collected, platform terms, API permissions, technical protection measures and internal review records. The more robustly those elements are documented now, the less likely a business will be left with only abstract explanations later.

Cross-border organisations should also begin putting rights clearance and data governance on the same risk map. Too often, brand teams, legal teams, procurement teams, technical teams and compliance teams still operate in separate lanes: trade marks are handled by one group, AI by another. But the signal coming from Canada and Singapore suggests that this separation is becoming the risk. A single project may be slowed in Canada by an older official mark while the same project creates exposure in Singapore or elsewhere because training-data origin, supplier model terms or access basis are poorly documented. The more effective operating model will increasingly be an integrated framework that simultaneously addresses clearance barriers, data provenance, contract responsibility and dispute jurisdiction.

Seen in that light, the Canadian and Singapore developments are not two isolated IP news items. They are a forward-looking warning about practice over the next several years: market entry, technology compliance and dispute readiness will increasingly turn on the same underlying capability—whether a business can turn its claims into a fact structure that others can inspect, question and test inside a real procedure.

This column is provided for general information only and does not constitute legal advice or a formal service recommendation. Specific matters should be assessed case by case and against the latest laws, examination practice, official notices and regulator views.