Five Takeaways from RegTech Summit London 2026

I recently attended RegTech Summit London expecting AI to dominate the conversation. It did, but I came away thinking about something broader: technology is beginning to challenge some of the assumptions on which traditional compliance has been built.

RegTech Summit London brings together senior leaders from financial institutions, regulators and technology providers to discuss how technology is reshaping compliance across European capital markets. Now in its tenth year, the summit attracts practitioners across compliance, risk, data, technology, regulatory reporting and financial crime, alongside the wider RegTech ecosystem

One audience poll captured the position well. 50% of respondents said their organisation was already scaling AI into production workflows, while 35% were running pilots. Yet only 3% considered AI to be business as usual in compliance.

AI has moved beyond experimentation, but the operating models around it are still catching up. These were the five themes that stayed with me.

1. T+1 looks very different when markets don’t stop

The move towards 24/7 markets has implications beyond trading infrastructure. It also challenges compliance processes designed around a world with clearer boundaries between one trading day and the next.

One presentation illustrated this particularly well. T+2 provided two overnight windows in which controls could run. T+1 reduces that to one. Move towards atomic settlement and 24/7 markets and there may be no overnight batch window at all.

Those windows currently accommodate processes such as reconciliations, alert reviews, sanctions rescreening and exception management. As the window disappears, some of those controls will need to operate very differently.

This does not mean every compliance decision needs to happen instantly. It does mean the gap between activity taking place and surveillance understanding it is likely to get smaller.

Always on compliance is therefore about more than doing the same surveillance faster. It increasingly means identifying and understanding risk while the surrounding context is still relevant.

2. Surveillance is where AI could have one of its biggest impacts

That change becomes particularly interesting when applied to surveillance.

In another audience poll, 57% thought compliance surveillance and monitoring was the role most likely to be redefined as AI adoption progresses.

Current adoption gives some indication of where that change is starting. 40% of respondents said they were using AI or machine learning for alert generation and detection, with another 27% using it for alert triage and prioritisation.

More advanced applications were much less common. Only 7% were using it to connect trade and communications data and 7% for behavioural or anomaly detection.

That gap is interesting because connecting different signals is where some of the greatest potential lies.

A trade may look unremarkable on its own. A conversation may contain nothing that triggers a traditional alert. Combined with surrounding activity, however, the picture can change.

AI does not necessarily need to replace existing rule based surveillance to deliver value. It can sit alongside those controls, helping interpret unstructured information, connect signals and give investigators better context around the activity that deserves their attention.

3. We are governing AI before we have agreed what AI means

Another theme from the day was how broadly the term AI is now being used.

Machine learning, generative AI, large language models, AI agents and agentic AI are regularly grouped together despite having very different capabilities.

That distinction becomes important in regulated environments. There is a considerable difference between a model classifying a communication, one generating a summary, and an agent capable of deciding what happens next and taking action.

Each creates different considerations around autonomy, accountability, explainability and human oversight.

As these technologies move into production, organisations need a common language for describing what they are deploying, what those systems are permitted to do and where responsibility sits. They also need ways to test and evidence that AI systems continue to behave as intended, something we explored in our guide to AI Assurance.

The fact that half of the audience said they were already scaling AI into production makes that need much more immediate.

4. AI could change how we think about data silos

For years, fragmented data has often led to another consolidation project: build a warehouse, create a data lake and bring everything together. The discussions at the summit challenged that assumption.

The audience itself reflected the reality of today’s infrastructure. 38% described their data as fragmented or siloed across systems and another 46% said it was only partially consolidated. Just 13% reported having a centralised source of truth.

What was more interesting was where attendees thought improvement would come from. When asked what would have the biggest impact on data consistency over the next two years, 42% selected AI driven data quality tools and 42% stronger governance and data ownership. Only 8% chose centralised data architecture.

That suggests a different way of thinking about the problem.

As AI becomes better at interpreting information across systems, firms may not need to physically consolidate everything before they can make sense of it. Trading information can remain in one system, communications in another and other compliance data elsewhere, with an intelligence layer helping connect the relevant information when required.

There is an important distinction between integrating the data and integrating the understanding of the data.

Governance, provenance and access controls remain essential. For communications compliance in particular, the archive can remain the authoritative source for the original record while AI helps connect that record with relevant information held elsewhere.

This is an approach we’re already exploring at Argus Archive through our Microsoft 365 Copilot integration. Rather than copying communications into another repository, governed communications held in Argus Archive can be accessed alongside information from other enterprise systems, while the original records, permissions and retention controls remain within the archive.

We explored this idea in more detail in why AI makes communications archives more important, particularly the role of the archive in reconstructing AI assisted decisions and demonstrating human oversight.

5. The elephant in the room: what comes after AI?

Financial institutions are investing heavily in adapting their infrastructure, governance and compliance processes to AI. Meanwhile, another technology challenge is already approaching: quantum computing.

One presentation framed the records problem as “harvest now, decrypt later.”

The principle is particularly relevant to financial services because many records created today must remain secure for years. Communications and order records can have retention periods of five years or more, while other regulatory and contractual records may need to be retained considerably longer.

Information encrypted securely today could potentially be collected and retained by an attacker with the intention of decrypting it once sufficiently capable technology becomes available.

For organisations responsible for long term records, quantum therefore isn’t solely a problem to address when quantum computers arrive. The longevity of the information means it can become a records management issue much earlier.

What struck me was the planning challenge. Firms are still working through the investment and governance required for AI while another potentially significant technology shift is already on the horizon.

AI will not be the last disruption compliance teams have to accommodate.

Building for what comes next

My biggest takeaway from RegTech Summit London was the pace at which the environment surrounding compliance is changing.

Markets are becoming more continuous. AI is already moving into production, but it is far from business as usual. Surveillance is beginning to move beyond isolated alerts towards richer context, while AI may also challenge long held assumptions about how data needs to be organised.

For communications compliance, the foundations remain important. Organisations need reliable records, clear governance and the ability to reconstruct what happened. But they may not need to centralise every piece of information before they can extract intelligence from it.

The archive can remain the authoritative communications record while AI provides a connective layer to the wider compliance environment. Quantum provides a useful reminder of why those foundations need to be built for the long term.

The challenge is bigger than moving from T+1 to always on compliance. It is building compliance infrastructure that can continue to adapt when the technology changes again.