FINMA and IOSCO Launch AI Tools for Crypto Market Supervision
Regulators are building the tools to watch crypto markets in real time, and the first phase launched in Zurich yesterday. Speaking at the Point Zero Forum on 23 June 2026, FINMA chair Marlene Amstad set out a clear picture: artificial intelligence is reshaping financial supervision just as decisively as it has reshaped financial markets themselves, and digital assets are the area where supervisory ambition most conspicuously outstrips current capability. The gap is closing. For accounting firms, auditors, and CFOs with digital asset exposure, the direction matters now, not when the tools go live.
What the IOSCO SupTech Survey Found
FINMA chairs the IOSCO SupTech Forum, a strategic initiative spanning more than 130 countries that together cover over 95 percent of the world's securities markets. A survey published the week of the forum, covering roughly three-quarters of global markets, mapped where supervisory technology currently sits and where it is heading.
Two goals, many use cases
All SupTech applications fall into one of two categories. The first is efficiency: doing familiar supervisory tasks faster and at lower cost. The second, and more significant, is effectiveness: generating insights that were simply not achievable before. The survey found that two-thirds of authorities are pursuing SupTech specifically for that deeper capability, the move from automation to what Amstad called "cognitive augmentation." Seeing what was previously invisible is the goal.
Use cases already in deployment or development include automated document analysis, market abuse investigations, sentiment analysis, and the monitoring of crypto exposures. That last item sits at the top of regulators' priority lists globally, and it is the one where stated interest most clearly runs ahead of actual deployed activity.
Digital assets: the leading edge
Amstad was direct about why digital assets dominate the SupTech agenda. The token world is expanding on two fronts simultaneously. Established crypto-assets like Bitcoin continue to grow in market importance. Stablecoins now have an estimated circulating supply above USD 300 billion, up tenfold from earlier baselines. Tokenisation of securities and other real-world assets is accelerating. Each of these trends generates on-chain data at a scale and speed that traditional supervisory processes cannot handle without technology.
Understanding how global stablecoin growth is reshaping accounting standards is one lens on the same pressure. The supervisory side is now moving in parallel.
The SupTech Sprint: What Happened in Zurich
The day before Amstad's address, FINMA hosted a hybrid SupTech Sprint in Zurich. More than 100 technology and policy specialists attended, representing regulators from jurisdictions covering roughly two-thirds of global securities markets by value. The event launched the first phase of building AI tools specifically designed for crypto-market supervision.
Scope and collaboration model
The Sprint model is deliberately collaborative. Supervisors worldwide face structurally identical problems when monitoring digital asset markets: pseudonymous addresses, cross-chain activity, fast-moving price discovery, and a data volume that dwarfs traditional securities markets. Because the challenge is shared, the solution development is shared too. FINMA's role as IOSCO SupTech Forum chair means Swiss regulatory thinking is directly shaping what these tools will look like when they eventually face live markets.
That international coordination dimension matters for firms operating across multiple jurisdictions. A supervisory tool built with input from two-thirds of global securities regulators will, over time, establish de facto standards for what constitutes adequate crypto-market data and audit trail quality.
How FINMA Is Already Using AI Internally
Amstad gave two concrete examples from FINMA's own operations, both of which illustrate principles that will carry through to how supervised firms are eventually assessed.
Pre-inspection document review
Before on-site inspections, the volume of documents FINMA staff must review has grown substantially. The authority is developing a generative AI tool that reads this material and flags passages worth closer attention, drawing on patterns identified across a large number of past inspections. The system runs on FINMA's local infrastructure and is currently at proof-of-concept stage.
Critically, the architecture uses a two-model design: one model proposes anomalies, a second independent model checks each suggestion against the source text to filter out hallucinations. Only suggestions that survive that second pass reach a human supervisor. The supervisor then makes the judgment call. Responsibility does not transfer to the algorithm. Amstad described this not as a technical detail but as a principle.
Insider trading detection and network analysis
The second example is more ambitious. Detecting insider trading requires two linked tasks: spotting suspicious trading patterns in transaction data ahead of a significant price move, and then demonstrating that those trades were informed by material non-public information, tracing how it flowed through a network of relationships. FINMA is developing AI tools using supervised machine learning models and network analysis to support both tasks. The approach applies equally to crypto-asset markets as to traditional securities, given that on-chain transaction data is, in principle, more complete and machine-readable than many legacy data sources.
Three Systemic Risks Regulators Are Watching
Alongside the operational SupTech work, Amstad identified three structural risks that AI introduces to financial markets. All three have direct compliance implications for firms.
Speed and the shrinking error window
Price discovery that once took minutes now happens in milliseconds. Credit decisions that took days can be settled in seconds. The efficiency is real, but the window in which an error can be caught before it propagates has narrowed correspondingly. For crypto markets, which already operate around the clock with no circuit breakers equivalent to those on traditional exchanges, this compression is acute.
Concentration in model providers
The most capable AI models, and the computing infrastructure behind them, are concentrated among a small number of providers. If a significant portion of the financial industry relies on the same few models for risk scoring or fraud detection, a single outage or a single flawed model output becomes a potential systemic event rather than one firm's operational problem. Regulators are paying close attention to third-party dependencies of exactly this kind.
Governance and accountability
Accountability cannot be delegated to an algorithm. That is the regulatory principle Amstad articulated most clearly. Decisions must remain explainable, and responsibility must stay with identifiable humans. This has direct implications for how firms document AI-assisted compliance decisions, including those involving digital asset transaction monitoring and AML screening. The expectation is not that firms avoid AI, but that they deploy it within a governance framework that keeps humans responsible for outcomes.
The governance challenge is equally visible in the AML context. For background on how regulators are approaching AI-enhanced fraud at the criminal end of the spectrum, the Interpol data referenced in Amstad's address is instructive: AI-enhanced fraud is described as four-and-a-half times more profitable than traditional methods, with agentic systems now capable of running entire fraud campaigns autonomously. Interpol calls this the industrialisation of fraud. SupTech is, in part, a direct response to that industrialisation.
For a related look at what the EBA's digital currency report means for AML and licensing, the supervisory trajectory at European level is consistent with what FINMA and IOSCO are building toward globally.
Practical Implications for Accounting Firms and CFOs
The SupTech Sprint and the IOSCO survey are not abstract policy exercises. They will shape what regulators expect to find when they examine firms with digital asset exposure, and they will do so sooner than many compliance calendars currently assume.
Data quality and audit-trail requirements
AI-powered supervisory tools are only as good as the data they ingest. When regulators develop market surveillance tools calibrated to on-chain and exchange-level transaction data, they are implicitly setting a floor for what firms must be able to produce. An accounting function that cannot reconstruct the full lifecycle of a crypto position, including wallet addresses, counterparty identifiers, timestamps to the second, and fair-value movements, will find itself exposed when a supervisory query arrives.
Robust crypto bookkeeping software that captures this data at source, rather than retrospectively reconstructing it from exchange CSV exports, is the practical starting point. The same structured data that feeds a tax calculation also feeds an AML review and, increasingly, a supervisory data request. Firms that treat these as separate workflows are building fragility into their compliance posture.
AML program calibration
The specific use cases FINMA named, crypto exposure monitoring, insider trading detection, and network analysis, tell compliance officers something concrete about what supervisory scrutiny looks like in the near term. AML programs that rely on manual transaction review or periodic batch screening are structurally mismatched with the real-time, network-aware tools regulators are building. The question for a compliance team is not whether to introduce technology-assisted monitoring, but how quickly and with what governance wrapper.
Switzerland-specific considerations
For firms operating under FINMA's jurisdiction, or for international firms with Swiss group entities, the domestic dimension is direct. FINMA has run three AI surveys of the Swiss financial industry and the findings show adoption is no longer siloed by function. Use cases now span the entire value chain. FINMA's own AI deployments, still at proof-of-concept stage, will mature. When they do, the evidence gathered during an AI-assisted pre-inspection review will have been shaped by training data from past inspections. Firms that have previously been examined will, in effect, have contributed to that training set. Consistent, well-documented compliance records across periods are not just good practice; they are a form of institutional memory that interacts with supervisory AI whether firms intend it or not.
Frequently Asked Questions
What is SupTech and why does it matter for crypto firms?
SupTech is the use of technology by regulators to modernise financial supervision. For crypto firms, it matters because regulators are building AI tools specifically to monitor digital asset markets, detect suspicious trading patterns, and analyse on-chain data. As these tools mature, the data quality and audit-trail standards they require will effectively become compliance expectations for regulated entities.
What did FINMA actually launch at the Point Zero Forum?
FINMA, in its role as chair of the IOSCO SupTech Forum, co-hosted a SupTech Sprint in Zurich on 22 June 2026. The event brought together more than 100 specialists from jurisdictions representing roughly two-thirds of global securities markets by value. The Sprint launched the first phase of developing AI-powered tools for crypto-market supervision, covering areas such as market abuse detection and crypto exposure monitoring.
How does the two-thirds-of-global-markets figure affect firms outside Switzerland?
Because the IOSCO SupTech Forum spans over 130 countries covering more than 95 percent of global securities markets, the tools and standards developed through the initiative are designed for cross-border applicability. A firm regulated in any participating jurisdiction, which includes most major financial centres, should expect that the supervisory data expectations emerging from this work will eventually reach their home regulator.
What does FINMA's human-in-the-loop principle mean for compliance teams?
FINMA's position is that AI tools must support human judgment, not replace it. For compliance teams, this translates directly: any AI-assisted AML screening, transaction monitoring, or risk-scoring process must be designed so that a named individual remains accountable for each decision. Documentation of the human review step is not optional. Regulators will look for it.
Should firms update their digital asset accounting software in response to this development?
The SupTech initiative reinforces what good practice already requires: granular, structured, real-time transaction records for all digital asset positions. Firms using digital asset accounting software that captures wallet-level data, timestamps, counterparty identifiers, and fair-value movements at source will be better placed to respond to supervisory data requests generated by AI-powered surveillance tools. Firms relying on manual reconciliation or periodic exports should treat this as a prompt to review their data architecture.
Source: FINMA