AI has come a long way, baby.
In a very short amount of time, too.
And it’s not just the technology that we’re talking about, but the regulatory discussions around artificial intelligence in the financial services industry. Just 12 short months ago, it seems like much of the regulatory debate around AI and finance revolved around whether AI should be integrated and deployed at all within financial processes and workflows.
Around the world, we’ve blown way past that debate—so quickly that regulators are starting to notice that the technology is evolving and being used within finance faster than they can address its use.
We’re over our skis.
As a result, U.S. financial regulators have quietly coalesced around one philosophy: artificial intelligence is becoming an accepted component of banking and financial services, but its adoption must occur within existing frameworks governing safety and soundness, consumer protection, fiduciary responsibility, fair lending, operational resilience, cybersecurity and model governance—which, of course, raises many questions regarding whether existing frameworks can be applied to AI.
Where the Fed Stands
On May 27, Governor Lisa Cook delivered a major speech examining AI’s opportunities and risks for the economy and financial system. Rather than portraying artificial intelligence solely as a technological disruption, Cook described it as a broad economic force capable of improving productivity, accelerating innovation and reshaping labor markets while simultaneously introducing new concentrations of operational and financial risk. She emphasized that policymakers must understand both dimensions simultaneously rather than focusing exclusively on either innovation or regulation.
Cook’s remarks reflected a broader theme emerging across federal agencies: AI should neither be viewed as an existential threat requiring immediate prohibition nor as a technological miracle exempt from oversight. Instead, regulators increasingly describe AI as another important financial technology requiring disciplined risk management.
On July 14, Federal Reserve Vice Chair for Supervision Michelle Bowman delivered remarks focused on responsible innovation and financial inclusion. Bowman argued that AI has substantial potential to expand financial inclusion by helping banks better evaluate borrowers who traditionally lack extensive credit histories. Using broader datasets and more sophisticated analytical techniques, AI could increase access to credit among underserved populations.
Yet Bowman immediately paired that optimism with an equally important caution. Credit decisions directly affecting consumers raise significant legal and regulatory issues involving fair lending laws, consumer protection requirements and anti-discrimination standards. AI may expand access to credit, but institutions remain responsible for ensuring automated decisions comply with existing law.
AI & Fiduciary Duty, Part One
One of the most important themes emerging from regulators—although often indirectly—is that artificial intelligence does not reduce fiduciary responsibility. As AI increasingly assists portfolio construction, financial planning, investment research and client communications, questions naturally arise regarding legal responsibility when algorithms produce flawed recommendations.
Federal officials have not suggested AI may replace fiduciary judgment. Instead, regulators consistently emphasize governance, oversight and accountability. Human decision-makers remain responsible for ensuring AI-generated outputs satisfy applicable legal standards. That philosophy has significant implications for financial advisers experimenting with AI-powered planning tools, portfolio optimization engines and autonomous financial agents.
An adviser cannot simply defend an unsuitable recommendation by claiming the software produced it. Likewise, banks cannot avoid responsibility for discriminatory lending outcomes by attributing them to machine-learning models. Regulators continue to expect explainability, documentation and human accountability, particularly where consumers experience meaningful financial consequences.
AI & Fiduciary Duty, Part Two
For wealth managers and registered investment advisers, perhaps the most consequential regulatory issue emerging this summer concerns the relationship between AI-generated recommendations and fiduciary obligations.
The fiduciary standard has always required advisers to place clients’ interests ahead of their own, exercise reasonable care and diligence, disclose conflicts of interest and make recommendations appropriate to each client’s circumstances. None of those obligations disappear simply because artificial intelligence becomes involved in the decision-making process.
Indeed, many legal experts argue the opposite may prove true: firms using AI could ultimately face higher expectations because they possess increasingly sophisticated analytical capabilities. If AI systems can identify conflicts, improve portfolio construction or detect unsuitable recommendations more effectively than traditional methods, regulators may eventually expect firms to incorporate those capabilities into their supervisory processes.
This possibility represents an important shift in thinking. Artificial intelligence may evolve from being viewed as an optional productivity tool into becoming part of what regulators consider prudent business practice.
At the same time, AI introduces new fiduciary risks. Generative models can hallucinate facts, produce fabricated citations, misunderstand client objectives or generate recommendations that appear persuasive but lack sound analytical foundations. Agentic systems could compound these risks by executing transactions before human supervisors have an opportunity to intervene.
Bipartisan Concerns in Congress
Although federal banking regulators have largely relied on supervisory guidance rather than new rulemaking, Congress has become increasingly active in examining artificial intelligence’s implications for financial services. Hearings held during the reporting period demonstrated growing bipartisan interest in understanding not simply generative AI, but the next generation of autonomous—or “agentic”—AI systems capable of initiating financial transactions, negotiating contracts and making decisions with limited human intervention. Congressional committees overseeing financial services have questioned regulators, technology executives and industry representatives about whether existing statutory authorities are sufficient to address AI-driven financial markets or whether new legislation will ultimately be required.
Much of the discussion has moved beyond familiar concerns about chatbots and large language models. Instead, lawmakers increasingly focused on AI systems that could independently execute trades, open financial accounts, originate loans, negotiate insurance claims or provide personalized financial guidance. These developments raise questions about licensing, liability and fiduciary responsibility that existing securities and banking laws were never designed to answer.
One recurring question is deceptively simple: Who is legally responsible when an AI agent acts on behalf of a customer? Current law generally assumes that a human adviser, registered representative or financial institution remains accountable for recommendations and transactions. Yet as AI agents become capable of acting with increasing autonomy, determining responsibility becomes more complicated. Is liability assigned to the financial institution deploying the AI? The software developer? The cloud provider? The customer who authorized the AI’s operation? Or all of the above?
What About the U.S. Treasury?
When it wasn’t busy putting a sitting president’s likeness on a dollar coin, the U.S. Treasury Department likewise broadened its discussion of artificial intelligence during the reporting period.
Treasury’s public statements increasingly describe AI not merely as a financial technology issue but as one intersecting with economic competitiveness, national security, cybersecurity and financial resilience. The department has continued coordinating across federal agencies to evaluate AI’s impact on payment systems, financial-market infrastructure, sanctions enforcement, fraud detection and cyber defense.
This broader framing reflects a recognition that AI’s influence extends well beyond individual financial institutions. Treasury officials increasingly view artificial intelligence as critical infrastructure whose reliability and security could affect the stability of the broader financial system.
For example, if multiple major financial institutions rely upon the same foundation models or cloud-based AI providers, operational failures could become highly correlated. A significant disruption affecting one widely used AI platform might simultaneously impair banks, broker-dealers, payment processors and asset managers. Such concentration risk has become one of the dominant themes across federal regulatory discussions. Rather than asking whether AI models are accurate, policymakers increasingly ask whether too many institutions depend upon the same underlying technology providers.
ChatGPT’s 5 Trends in Financial AI Regulation
First, regulators have moved beyond debating whether financial institutions should adopt AI. The discussion now centers on governing inevitable adoption responsibly.
Second, fiduciary responsibility remains firmly attached to regulated firms and licensed professionals. Artificial intelligence may assist decision-making, but regulators consistently reject the notion that software itself can bear legal responsibility.
Third, governance has become the defining regulatory concept. Across banking, securities, insurance and fintech, supervisors increasingly evaluate documentation, oversight, validation, explainability and accountability rather than focusing exclusively on technical performance.
Fourth, international regulators are converging around remarkably similar supervisory principles despite differing legal systems. Human oversight, operational resilience, cybersecurity, transparency and proportional risk management now appear repeatedly in guidance from the Federal Reserve, Bank of England, ECB, BIS, IMF, OECD and Financial Stability Board.
Fifth, autonomous AI agents—not today’s generative chatbots—represent the next major regulatory frontier. Existing financial law largely assumes human decision-makers remain at the center of regulated activities. Agentic systems increasingly challenge that assumption.
Heading Off a Crisis in Confidence
What concerns regulators is not artificial intelligence itself. It is the possibility that institutions deploy increasingly autonomous systems without preserving the accountability, transparency and public confidence upon which financial markets ultimately depend.
For decades following the global financial crisis, financial regulation concentrated on capital adequacy, liquidity, stress testing and systemic resilience. Artificial intelligence does not replace those priorities. Instead, it introduces a new layer of governance concerned with the quality of automated decisions, the concentration of technological dependencies, the resilience of digital infrastructure and the continuing responsibility of human professionals.
Competitive success in the AI era will not be determined solely by access to the most advanced models or the largest computing infrastructure. It will increasingly depend upon demonstrating that AI systems operate within governance frameworks capable of satisfying regulators, protecting consumers and maintaining public trust.
That may ultimately become the defining characteristic of financial AI regulation. Rather than serving as an obstacle to innovation, regulation is evolving into the mechanism through which confidence in AI-enabled financial services is established. Institutions able to combine technological sophistication with disciplined governance are likely to enjoy not only regulatory credibility but also a durable competitive advantage in the next generation of financial services.
Researched by DWN Staff
