William Wright
This article was produced in collaboration with Tech Trek Events for the NexGen Banking Summit 2025.
Calls for banking deregulation, embraced by the Trump Administration, poses a fundamental challenge to the post-2008 regulatory frameworks.
Combined with global divergence in AI regulation, the banking industry needs cross-jurisdictional platforms like the NexGen Banking Summit 2025 to unite and create sufficient AI strategies.
With senior leaders from institutions like JP Morgan Chase, the Bank of England, HSBC, and Citi taking the stage, the summit presents a global perspective on GenAI in banking, collectively encouraging greater compression of what GenAI means for their industry.
Regulation can be a double-edged sword; it gives and it takes. Well-designed, well-enforced regulation creates a stable framework of general rules that maintains customer confidence and encourages innovation. But regulation can also act as a drag to innovation and efficiency, increase costs, and create even more uncertainty when inconsistently enforced.
What does this mean for AI innovation? What does this mean for AI safety? What does this mean for GenAI in banking?
Both banking and AI regulations are contentious issues. For banking, a ruleset produced in the wake of the 2008 financial crisis has paired with slow recovery, and the pursuit of economic growth has led many governments to consider deregulation, chief amongst these being the US.
Unlike banking, AI does not have a globally-accepted, formal framework to regulate from. Despite the work of numerous committees (such as the UN’s AI Advisory Body, the CAI, the OECD etc), AI regulation is fractured and the precedents are set by individual states.
AI is contentious because this lack of a globally-accepted framework means the basic principles – good and bad, moral and immoral, safe and unsafe – are hazy.
I recently attended the CII’s ‘Shaping the Future of Insurance’ where Professor Richard Susskind, author of ‘How to Think About AI’ and advisor to governments, presented the possible outcomes of AI.
These ranged from: ‘the technology is at its peak now (unlikely) and the impact will be tiny’, to ‘we will see continuous, exponential improvements over the next decade, the singularity is coming and humanity will be a footnote in the history of machines’.
This variance is significant because it highlights the lack of consensus about AI right now, and the lack of common ground means it is nearly impossible to regulate.
But AI regulation is seen as a priority for governments right now. With AI adoption reportedly offering an additional 15% in global economic output by 2035, and GenAI in banking alone offering as much as $340 billion a year in additional value, AI has been considered the solution to the downward trend in labour productivity in the major industrialised economies.
The increasing differences in global economic and political ideologies can quite interesting be captured in the debates on banking and AI regulation.
What this article will highlight is that despite individual nation states driving precedents, and despite the lack of a globally-accepted consensus on these issues, we exist in a globalised world.
We must promote collaboration and debate to reach consensus, and for AI and banking, events like the NexGen Banking Summit provide that platform.
Banking Regulation and De-regulation
Zooming in on just banking, regulatory adherence is a large part of what instills trust and confidence in clients and customers. Since 2008, a rise in scrutiny and rigid rules was meant to ensure 2008 never happened again.
However, with slow recovery from 2008 and stringent rules allegedly constraining financial institutions, harming competitiveness, and costing too much for the government to supervise, and too much for especially smaller banks to comply with.
There has been increasingly vocal calls for the deregulation of banks, backtracking on the Basel III accord and Basel III Endgame.
What are the Basel Accords and Basel III Endgame?
In response to the bankruptcy and subsequent crisis of the German Herstatt Bank in 1974, central bank governors from the G10 formed the Basel Committee on Banking Supervision (BCBS). This committee provided a forum for global cooperation on banking supervisory matters.
The first product was Basel I in 1988 (enforced by law in 1992) which set minimum capital requirements for banks.
Basel II in 2004 (enforced by law in 2006-2008) intended to supersede Basel I with a ‘three-pillar’ system of minimum capital requirements, supervisory review, and market discipline. However, as its implementation crossed-over with the financial crisis, it failed.
Basel III in 2010 aimed to correct the critical failures of Basel II by significantly raising the quantity and quality of required bank capital, introducing strict liquidity and leverage ratios, and imposing additional surcharges on globally systemically important banks (G-SIBs).
Basel III was the manifestation of crisis fear, and as the implementation deadlines rolled in for capital and liquidity requirements and leverage ratios, critics began to protest.
When Basel III Endgame’s (the third and final set of proposals for Basel III) 2025 compliance date approached, the Fed agreed to revise capital requirements. This has morphed into a greater movement within banking, with many large US firms calling for a general de-regulation.
How does this link to US AI regulation and AI in banking?
The general tone towards banking deregulation, set largely by large US banks and the Trump Administration, also maps onto the general outlook for US AI regulation.
Similar to Trump’s rejection of Biden’s banking regulation measures, Trump reversed Biden’s 2023 Executive Order 14110, “Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence”.
Replaced with Executive Order 14179 “Removing Barriers to American Leadership in Artificial Intelligence”, the current administration’s prevailing strategy is to foster rapid, uninhibited innovation to maintain global leadership.
“America’s AI Action Plan” prioritises removing regulatory barriers and promoting development, reflecting a belief that a light touch will spur a competitive edge. This has led to a reluctance in Congress to pass comprehensive, restrictive legislation, favoring voluntary guidelines instead.
Federal VS State Banking and AI Regulation
In the US, banking is regulated at both the federal and state level. Banks can be chartered at either the state or federal level, intended to promote a diverse ecosystem of smaller, regional institutions and larger, global banks.
AI is currently not regulated federally, and the states have been entrusted to set the safety guardrails. Certain states like Colorado and California are leading AI legislation, others are taking the federal ‘laissez-faire’ approach.
So while the US appears to be the central destination for a lax AI regulatory regime, fragmentation in the AI regulatory landscape makes it complicated for banks to meaningfully implement AI across multiple jurisdictions.
Comparison with Europe and the UK
This issue is further exacerbated by the difference in the US regulatory approach with the EU and the UK.
Take the UK, where banking deregulation has similarly grown in popularity reflected in government rhetoric (see Reeves on the end of ‘ringfencing’).
Certain narratives, such as those presented by Revolut in its three-year struggle to gain a full banking license or attacks on the PSR’s interchange limits, highlight how intense regulatory regimes stifle innovation, competition, and growth.
These narratives have been paired with AI regulation, with firms such as Google calling on the UK to adopt a ‘pro-innovation’ approach when regulating AI.
But the UK has also promised to adopt a principles-approach that maintains safety and security, and the specific of what this means are vague.
For Europe, the EU AI Act in August was the world’s first comprehensive AI law. Compared to the UK’s current approach, and especially the US’ laissez-faire approach, the EU has set itself apart on the matter of AI regulation.
The AI Act aims to incentivise trustworthy AI by categorising systems based on risk and establishing rules for development, deployment, and use. The regulation bans unacceptable-risk AI like social scoring and manipulation, sets strict requirements for high-risk AI (e.g. banking), and mandates transparency for systems like deepfakes.
Regulatory Divergence: What About the Future?
These two debates on banking and AI regulation are intimately connected to a future that will likely harness the power of GenAI.
The world’s largest banks are now caught in a tricky regulatory bind. While the US champions a deregulatory approach that fragments AI standards across individual states, the EU implements the sweeping, comprehensive guardrails of the EU AI Act. The UK acts as an intermediary, waiting for a comprehensive framework that combines pro-innovation and safety.
Since no government can agree on what constitutes “safe” or “prudent” innovation, global institutions are paralysed by compliance risk. We are not facing a question of if GenAI can deliver $340 billion in value, but how to deploy it without risking a systemic collapse or a catastrophic fine.
Benefit of NexGen Banking Summit 2025
This is why events like the NexGen Banking Summit are so valuable; they serve as necessary cross-jurisdictional collaboration hubs for multinational financial institutions that must comply with all three regimes simultaneously.
The industry must stage collective discussions to understand the impact of GenAI, and strategise how to navigate the tricky regulatory future ahead.
This week, banking leaders will move beyond theory to address the tactical challenge of risk management, model transparency, and auditability, allowing banks to grow their GenAI advantage without jeopardising regulatory compliance.
With demand already high and limited seats available, now is the best time to secure your place.
This article is the second in the NexGen Banking Summit Series. Stay tuned for the next piece which will highlight the key takeaways of the conference.
Featured image via FreePik.
