Karsten Wenzlaff, Advisor
August 26th, 2025
Feb 4, 2026 | NCFA Expert Intelligence | AI Risk and Governance

On February 3 2026, Dentons published a Voice from the Market North American Legal AI report following the firm’s inaugural Legal AI Summit. The report captures data insights from legal, business, and operations leaders across sectors about how organizations are handling real world AI deployment, specifically where AI adoption is creating legal and governance risk before regulators provide clear framework.
The findings matter for fintech founders and operators because legal teams are often the first to see where risk accumulates, such as in contracts, vendor relationships, employee use, and product liability, long before the market or regulators respond.
Most organizations in the survey reported they are designing their own AI governance frameworks because universal regulations do not exist yet. That gap puts legal risk in the critical path for AI use inside financial products and services.
“We want to cover the basics, but we don’t want to stymie innovation or scare people away with too many ‘do nots.’”
This comment above is from a survey participant highlighting the tension leaders face: Smart governance is needed, but overly prescriptive rules can impede experimentation and execution. So it's a real operational conflict that stakeholders must resolve long before policy catches up.
The governance gaps in the report align with real risks fintechs already encounter:
Many fintechs are building AI capability faster than they are formalizing guardrails, which can expose firms to legal, operational, and compliance risk across functions where AI is integrated in workflows. Whether that's in credit decisions, document review, marketing, fraud monitoring, or customer support.
It also intersects with other areas NCFA has covered. For example, recent regulatory signals from the CSA–CIRO joint finfluencer guidance and the rise of stablecoin policy frameworks both show that Canada’s regulatory perimeter is tightening around behaviour and risk, not just product categories.
If legal risk surfaces first in AI contracts, vendor relationships, and internal use, where should fintechs build their earliest governance guardrails: in contracts, in internal policy, or in product design?
Regulation typically always lags innovation. Legal and governance frameworks are being created inside enterprises today because no universal rulebook exists. Fintech innovators and operators who anticipate these gaps and embed guardrails early will be better positioned to scale responsibly and avoid costly legal exposure down the road.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada’s Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
![]() | ![]() | ![]() |
|---|---|---|
![]() | ![]() | ![]() |
Jan 29, 2026 | NCFA Fintech Market Activity | Insurtech and Automation
Image: Freepik/rawpixel.com
On January 29 2026, Manulife Canada announced changes to its life insurance application process that allow eligible applicants to receive automated approvals in as little as two minutes.
The update combines a redesigned digital application with enhancements to Manulife’s AI driven underwriting engine, MAUDE. According to the release, the new workflow eliminates many manual steps that previously added days or weeks to approval timelines for straightforward policies.
Manulife reports that the updated process now supports instant decisions for a broader set of low risk applicants. In cases that still require review, the system routes applications more efficiently by pre screening risk factors before human involvement.
The operational impact is material. Life insurance underwriting traditionally relies on paper based forms, medical questionnaires, and extended back and forth with applicants. Compressing approvals to minutes reduces abandonment, shortens sales cycles, and lowers per policy acquisition costs.
The release also highlights that Manulife’s digital life insurance business has grown steadily as customers shift toward online purchasing. Faster approvals support higher conversion rates while reducing pressure on underwriting teams as volumes scale.
This is not a cosmetic change. Automated underwriting at this speed requires confidence in data quality, model governance, and exception handling. Errors scale as fast as efficiency. That makes oversight, auditability, and fallback paths just as important as speed.
Karen Cutler, Chief Underwriter, Manulife Canada:
"Since launching last fall, advisor adoption has been strong. By December, more than half of eligible cases – 58% had approvals processed automatically through MAUDE, a 56% increase from pre-launch. By combining smarter questions with advanced AI, we're delivering a faster, more intuitive experience – without compromising on quality or protection."
If two minute approvals become the baseline for simple policies, where do insurers deliberately slow the process to manage risk, trust, and regulatory expectations?
Manulife’s update shows how artificial intelligence in insurance has moved past pilots. The competitive gap now sits between firms that embed automation directly into core underwriting and those still layering digital tools on top of legacy processes.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer to peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
![]() | ![]() | ![]() |
|---|---|---|
![]() | ![]() | ![]() |
AI Governance and Risks | Jan 23, 2026

Image: Freepik/DC Studio
On January 21, 2026, Anthropic announced a new 2026 constitution for Claude, an internal framework that guides how its AI models reasons and responds, moving it's intelligence beyond guardrails and towards explicit value-based judgment, see: Claude's new constitution. This update builds on Anthropic’s original 2023 constitutional AI, which focused on constraining harmful outputs instead of embedding decision logic inside the model.
The timing matters. Policymakers are already warning that AI use in financial services is creating risks that existing oversight struggles to address, including opaque decisioning, unclear accountability, and consumer harm, as detailed in the UK Parliament Treasury Committee’s report on artificial intelligence in financial services. In Canada, the federal government launched a Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, setting expectations for governance, oversight, and risk management at the organizational level rather than inside individual models. Since its launch in September 2023, a growing group of technology, financial services, and infrastructure firms have signed on with early alignment around common principles even as binding AI legislation is still under development.
Seen in that context, Claude’s 2026 constitution provides a clear example of how decision making is being built into AI models themselves, while responsibility for outcomes still sits with the institutions that use them.
2023 constitution
The original constitution operated primarily as a set of behavioural constraints. It instructed the model on what to avoid and what to prioritize, largely to prevent harmful or non compliant outputs. The emphasis was containment.
2026 revision
The revised constitution introduces embedded judgment logic. Instead of checking outputs against rules, the model is trained to reason through competing objectives using an explicit hierarchy of values.
Under the 2023 framework, a customer support assistant helping with a disputed transaction would follow predefined rules. If the request met policy, the model would proceed. Under the 2026 framework, the model may pause or redirect even when rules allow action, if it determines that safety or downstream harm outweighs helpfulness. The outcome changes not because policy changed, but because judgment moved inside the model. The practical result is that product teams can no longer assume consistent behaviour across similar cases, since internal judgment may lead to different responses even when inputs look the same.
2023 constitution
Values such as safety, helpfulness, and compliance were present but not clearly ordered. The model generally tried to satisfy all simultaneously.
2026 revision
Anthropic now makes the hierarchy explicit. Broad safety and ethics take priority, followed by compliance with rules and policies. Helpfulness ranks last.
Consider a compliant onboarding flow for a higher risk customer segment. In 2023, the model would guide the user through required steps as long as eligibility criteria were met. In 2026, the model may slow the process, add cautionary language, or decline to provide certain guidance if it determines that risk or misuse concerns outweigh the value of completing onboarding smoothly. Over time, this can change conversion and completion rates in ways that are difficult to trace back to a single rule or policy change.
2023 constitution
Explainability centered on technical or policy based reasoning. Outputs were justified by reference to rules or restrictions.
2026 revision
The model is encouraged to explain decisions using value based reasoning, such as safety considerations or ethical caution.
In a credit related inquiry, a 2023 style response would reference eligibility thresholds or policy rules. A 2026 style response may explain that certain guidance cannot be provided because of safety concerns or potential harm. While that explanation may feel reasonable to a customer, it doesn't currently map cleanly to regulatory expectations that rely on objective, measurable criteria. This creates added pressure on compliance teams to reconcile human readable explanations with audit and supervisory requirements.
2023 constitution
Uncertainty was handled implicitly through restrictions.
2026 revision
Anthropic explicitly states that uncertainty warrants caution. The model is trained to err toward conservative responses when intent, impact, or downstream effects are unclear.
In financial education or product comparison tools, a 2023 model would provide general guidance within allowed boundaries. A 2026 model may decline to offer comparisons or suggestions if it believes the information could be misused or misunderstood. This reduces risk, but if left unmanaged it can gradually narrow what the product is willing to do, even when the underlying business intent remains unchanged.
Model risk no longer stops at accuracy, bias, or drift. Those risks are familiar. Since the 2026 constitution itself becomes part of the model, fintechs now inherit a set of embedded priorities that influence how decisions are made when rules alone are not enough. If those priorities are not understood and tested, teams can see unexpected behaviour even when inputs and policies stay the same.
This puts real pressure on procurement and vendor governance. Evaluating models based on performance and compliance checklists is no longer sufficient. Fintechs need to understand how a model handles tradeoffs, when it decides to be cautious, and when it may refuse or redirect actions that the business considers acceptable. Those behaviours directly affect customer experience, conversion, and operational consistency.
Accountability still sits with the organization. Even when an AI model makes its own call, regulators and customers still look to the firm using it. If an AI driven interaction leads to confusion, exclusion, or harm, the explanation and the responsibility is still firmly with the fintech provider, not the AI system or tool. Embedded judgment raises the stakes for internal oversight because there is less room to say the system simply followed a rule.
Canada’s voluntary AI code is an important step but it operates at a different level. It sets expectations for governance, monitoring, and oversight by organizations. It doesn't however address how a model reasons in real situations. That gap becomes more important now that judgment is being built into some models directly. Fintechs can be in a situation where the code meets expectations but they still face questions that they aren't fully prepared to answer when an AI behaves in ways that are technically compliant but operationally unexpected.
The change from Claude’s 2023 constitution to the 2026 update is a breadcrumb trail showing where AI governance is heading. For many firms there could be an emerging gap that shows up when a customer gets a different answer than expected, when a compliance team has to explain why an AI responded the way it did, or when a board asks how much control the firm really has over a third party model.
The real issue is not whether AI follows rules. It is whether fintechs understand how these systems behave in real situations, before that behaviour affects customers, regulators, or the business.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
![]() | ![]() | ![]() |
|---|---|---|
![]() | ![]() | ![]() |
RPAA | January 20, 2026

Image: Freepik
On September 8 2025, the Bank of Canada published a letter to PSPs on a trust tax issue, that said safeguarding end-user funds in “trust” accounts, a requirement under the Retail Payment Activities Act (RPAA), could create unintended tax obligations under the Income Tax Act. The impact threatened to burden PSPs with additional tax filings and compliance complexity for simply holding user funds.
For fintech founders, investors and advisors building or backing PSPs, safeguarding end-user funds is a core requirement of the RPAA. If standard trust arrangements triggered full trust tax treatment, PSPs could end up doing extra tax filings simply to meet payments rules. That would force teams to spend time and money on tax compliance instead of building products, serving customers, or growing the business.
On December 23 2025, the Bank of Canada issued an update that confirmed the Department of Finance proposed changes to the Income Tax Act so that RPAA safeguarding trust arrangements would not be treated as formal trusts for tax purposes. Instead, qualifying arrangements would be treated as deemed loans from end users to PSPs, removing the risk that PSPs would face extra trust tax filings requirements.
The change is intended to apply retroactively to September 8 2025, subject to legislative approval.
The lesson here is that staying engaged with regulators, raising issues early, and participating in public comment or industry groups can help outcomes that materially affect operations and capital efficiency.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
![]() | ![]() | ![]() |
|---|---|---|
![]() | ![]() | ![]() |
AI | January 20, 2026

Image: Freepik
On January 19, 2026 via Reuters, OpenAI's Chief Financial Officer Sarah Friar confirmed that its annualized revenue exceeded US $20B in 2025. Reported figures show revenue was $6B in 2024 and about $2B in 2023, meaning revenue more than tripled this past year alone, mirroring the scale of paid demand for OpenAI services.
Annualized revenue describes what a company would generate over twelve months if current sales levels stayed constant, and is a widely measure to assess momentum in subscription and usage-based businesses. For OpenAI, the number reflects sustained growth in paid ChatGPT subscriptions, enterprise contracts, and API usage across multiple industries.
OpenAI’s revenue growth closely follows its infrastructure expansion. Reuters reports that the company increased its computing capacity from roughly 0.6 gigawatts in 2024 to about 1.9 gigawatts in 2025, nearly tripling in one year. That level of scale requires access to capital, power, chips, and long-term supplier agreements that only a handful of firms can sustain.
As scale increases, it begins to set the rules of the market.
Infrastructure capacity now determines who can offer AI services at competitive prices, who can guarantee availability during peak demand, and who can absorb cost volatility without disrupting customers. Performance expectations increasingly reflect what the largest providers can deliver, not what smaller firms can economically support.
This trajectory does not mean startups disappear overnight, but it raises the survival bar.
Startups that rely on thin margins face the earliest pressure. This includes companies that resell model access, wrap generic AI features into existing products, or compete primarily on convenience rather than control over data or distribution. As OpenAI captures more value directly through subscriptions, enterprise contracts, and advertising, profit margins for intermediaries narrows.
The pressure shows up first in functions that scale with usage rather than revenue, such as customer support automation, fraud monitoring, transaction screening, onboarding workflows, and real time decisioning. In these areas, AI costs rise with activity even when revenue does not.
By contrast, startups with proprietary data, embedded customer relationships, or regulated market access retain leverage. The dividing line is no longer access to AI. It is control over economics, customers, and accountability.
AI pricing is typically usage based. As fintech products scale, AI expenses scale with them, often faster and with less predictability than revenue. This creates margin pressure, particularly in Canada where fintechs operate with smaller domestic scale, tighter capital markets, and higher regulatory expectations. What appears manageable during early deployment can become material at scale.
In regulated financial services, vendor dependency becomes operational risk, and operational risk becomes regulatory exposure.
There is also a governance reality. When AI influences credit decisions, disclosures, complaints, or customer outcomes, fintechs remain fully accountable even when they do not control model behaviour, pricing changes, or service limits.
Pricing power in AI is concentrating upstream. Intelligence is becoming a metered utility controlled by a small number of providers with the capital and infrastructure to set terms. Large global firms can absorb cost swings, negotiate volume discounts, and manage vendor risk across portfolios. Many Canadian fintechs cannot.
The strategic question is not whether to use AI, but how tightly to bind core operations to any single supplier. Fintechs that preserve provider flexibility, constrain AI use to high impact functions, and tie spending directly to measurable outcomes retain an advantage (for now). Those that treat AI as an invisible background feature risk discovering too late that it has become a fixed dependency with variable costs.
OpenAI’s $20B annualized revenue growth is about market structure limits. Scale now determines who sets prices, who absorbs volatility, and who carries risk. For Canadian fintech, AI decisions now belong alongside cloud infrastructure, payments rails, and core systems as board level considerations.The firms that stay competitive will be the ones that manage AI like any other core cost, with clear oversight and control.
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
![]() | ![]() | ![]() |
|---|---|---|
![]() | ![]() | ![]() |