ThetaRay after Sibos 2026: AI-native compliance moves from pilot to production

Brad Levy, CEO, and Garima Chaudhary, SVP Product  at ThetaRay, share their takeaways from Sibos 2026 in Miami, from the move to AI-native compliance to the company's plans for 2027.

Sibos 2026 took place in Miami from 28 September to 1 October, the first time the conference has been held in the city. ThetaRay, which provides AI-based financial crime compliance technology, attended with CEO Brad Levy and Garima Chaudhary, SVP Product. Shortly after the event, The Paypers spoke with both about what they heard from banks in Miami and where the company is heading next.

Instant payments expose the limits of static rules

For Chaudhary, the main takeaway from Sibos was the pressure instant payments place on compliance. Customers expect payments to settle quickly and safely, while most institutions still rely on batch processing and static rule-based monitoring. She estimated that more than 95% of the alerts such systems generate can be false positives, and said deepfakes and synthetic identities add to the strain.

Many of her conversations in Miami focused on replacing static rules with unsupervised machine learning models, which do not need labelled data and can adjust as products and customer behaviour change. Chaudhary noted that the approach is new to much of the industry but is ‘something we have been doing for 14 years’ at ThetaRay.

Her second takeaway concerned agentic AI. According to Chaudhary, the discussion has ‘moved from pilot to production’, and ThetaRay has customers at both stages. The focus is now on sustaining and scaling agentic AI with governance in place, so that humans stay in charge as adoption grows.

Levy agreed that the industry has moved on. ‘I think it's definitely beyond the pilot into production,’ he said. He attributed part of the forward-looking tone to the international audience at Sibos, noting that Europe has picked up AI-native technologies sooner, which he linked to the volume of cross-border activity and to a regulatory environment, including GDPR, that is demanding without limiting AI usage. A year or two ago, he said, many of the same people would have told him they were not ready.

Levy also described a change in how banks think about modernisation. Replacing core systems wholesale is impractical and takes too long, so institutions now modernise in smaller increments, and sometimes find that older components remain useful. He also observed more discussion of interconnectivity between traditional finance and blockchain-based infrastructure, and said the discussion stayed centred on payments efficiency and risk.

A missing compliance crowd

Levy, who described himself as new to Sibos and still relatively new to financial crime, said he went looking for compliance specialists in Miami and found fewer than he expected. Part of the reason was a compliance conference held in Las Vegas the same week, but he also saw a divide between the payments and banking community and the compliance community. In his view, ‘the compliance element was a little lost on Sibos’, and he expects compliance to become part of the mainstream payments conversation within the next couple of years, in the same way cyber security needs to be.

The typologies on banks' minds

Chaudhary said no single typology dominated her conversations at Sibos, although the cross-border use of shell accounts remains a priority for the institutions ThetaRay works with, particularly in relation to sanctions evasion. She described the actors behind these schemes as ‘the big fishes’, as opposed to individual offenders.

These schemes are difficult to detect for banks that rely on static rules and internal data, which is largely limited to customer account and transaction information. Mapping corporate structures, ultimate beneficial ownership, subsidiaries, and customers' customers requires more. Chaudhary said ThetaRay's unsupervised models draw on internal and external data, trace relationships across multiple hops, and look for anomalies against peer groups defined by country and line of business.

Levy pointed to sanctions and human trafficking as the topics drawing the most attention. Sanctions regimes shift with each new conflict, which means ‘set it and forget it type rules obviously don't make sense’. He noted that human trafficking falls mainly under AML transaction monitoring, while sanctions are largely a screening problem, and that clients increasingly want to address the two together. At country level, he added, each market is distinct because customer behaviour and national AML and banking laws differ.

Consumers are watching their banks

Asked whether consumer pressure is influencing financial crime spending, Levy said innovation now tends to flow from retail into institutional finance. Neobanks carry less technical and business-model legacy, and they are pushing traditional banks towards faster onboarding and real-time processing. ‘The retail consumer world is definitely leading tech innovation much more in the last decade than it ever has in the previous four decades,’ he said.

Chaudhary argued that consumers now apply the same ethical scrutiny to their banks that they apply to the clothes and food they buy. She cited the financial and reputational fallout banks face after enforcement action, including the USD 3 billion fine against TD Bank and the loss of valuation that followed. ‘Nobody wants to be in that type of news anymore,’ she said.

Levy linked this awareness to supply chains. The pandemic showed people how much depends on them, and he believes there is now wider recognition that ‘there's a money supply chain that is inextricably linked to actual supply chains’. With stablecoins and other digital assets, he sees value moving between fiat currency and other forms, including digital tokens, and he suggested that the term anti-money laundering is becoming too narrow. His proposed alternative is ‘AVL, anti-value laundering’.    

Real-time payments and fraud that cannot be prevented

On instant payments, Chaudhary was direct: ‘They're falling behind’. Combined with synthetic identities and deepfakes, faster payments widen the gap between what rule-based systems detect and what criminals do. Rules, she said, are designed around known patterns, such as a USD 10,000 transfer to a high-risk country, and criminals learn those thresholds. In her words, ‘rules can only find what is known to industry and known to organisation’. Unsupervised models can instead adapt to new behaviour and emerging risks. She added that fintechs have been more open to adopting this technology than larger banks, which remain more hesitant because of their legacy systems.

Levy described the trend as the ‘real-timeification of everything’ and said it requires a change in mindset. Institutions will not prevent every incident, so the priority becomes stopping an attack from repeating, with kill switches that can cut off a pattern quickly. Once funds have moved through several crypto wallets, he said, they are unlikely to be recovered. He also raised the question of liability, comparing it to the insurance debate around self-driving cars and to the way card networks allocate fraud losses. ‘We have work to do on architecture and on the liability of the risk when it does hit in size,’ he said.

Where the agent stops and the human decides

Describing how Ray, ThetaRay's agentic AI investigation suite, works in practice, Levy summarised the company's priorities for AI as ‘explainable, predictable, controllable’. ThetaRay has long exposed how its models reach their outputs. ‘We call it the glass box,’ he said, and he expects that transparency to make clients comfortable granting agents more autonomy sooner, although humans will still make the final call much of the time.

In Levy's account, the agent first reduces noise by clearing large volumes of false positives, then prioritises what remains and supports the investigation itself. Some cases can be almost fully automated and need only a final sign-off. Others involve an iterative process over several hours, with each step recorded in the audit trail. Clearing false positives and compressing investigations that take experienced staff hours deliver different kinds of value, and he does not expect the latter to become fully autonomous.

Chaudhary said agents take on the laborious part of the work. When an alert is raised, an investigator may need to review hundreds or thousands of transactions. Agents gather data from source systems, assess it against the institution's policies and red flags, make a recommendation and draft the case or SAR narrative, which is time-consuming and prone to inconsistency when done manually. ‘The authority is going to be with the humans,’ she said. Her suggestion is to ‘think of humans as managing a bunch of agents’.

Levy added one use he finds promising: the agent as ‘the tap on the shoulder to reprioritise your queue’, alerting an investigator to something that needs attention without the disruption of switching between systems.

ThetaRay's plans for 2027

Looking ahead, Levy said ThetaRay is focused on Europe, Latin America, Africa and the Middle East, with North America a major growth priority. ‘North America is particularly important because obviously the US is the biggest financial market in the world,’ he said, adding that maturing digital asset regulation will make the US and Canada more important still. The company has deliberately held back in Asia-Pacific for now. ‘We're overtly not spending time on APAC just as a business,’ Levy said, although he expects that to change in 2027.

On product, ThetaRay plans to offer its stack both as an integrated suite and as separate modules. ‘All of it will have an agentic flavour to it,’ Levy said. The company also expects to play a role in digital assets as traditional and decentralised finance converge, reaching that market through neobanks and other existing clients that want to support stablecoins on shared infrastructure. In his view, ‘the rule sets will be much more blended than people realise’.

Levy noted that many clients are already multi-jurisdictional. ThetaRay works with Santander and with I&M Bank in Kenya, among others, and a single client can mean deployment across several countries.

He also highlighted the changes to the team in 2026. Chaudhary was the first senior leadership hire after Levy, joining in March, and the company has since hired around 15 people, roughly 10 of them at leadership level, including hires from Oracle and Symphony AI. ‘We are specialists and we're betting on specialists,’ Levy said.

About the author

Vlad is a Senior Writer at The Paypers, working in the commercial team. He uses his research, content, and people skills for all editorial activities revolving around payments, banking, and fintech. Vlad has a degree in Biology and Molecular Genetics and an extensive background in creative writing. You can reach out to him on LinkedIn.

 

 

About the interviewees

Brad Levy is the CEO and board member of ThetaRay, where he leads the company’s mission to set the industry standard for AI-powered financial crime compliance. His three-decade career includes serving as CEO of Symphony, CEO of MarkitSERV and 18 years at Goldman Sachs, where he was a Managing Director and Global Head of the Principal Strategic Investments Group.

 

 

Garima Chaudhary is the SVP Product at ThetaRay. She previously served as the Global Head of Financial Crimes & Compliance Management Solution Engineering at Oracle, where she led a global team of experts in AML, KYC, sanctions, fraud, and advanced analytics. Garima is dedicated to harnessing AI, data, and cloud innovation to build scalable, future-ready defenses against financial crime.      

 

About ThetaRay

ThetaRay provides the end-to-end financial crime risk decisioning infrastructure for enterprises. Powered by advanced machine learning, ThetaRay integrates across the full customer lifecycle, from screening and customer due diligence to transaction monitoring. Trusted by global banks, fintechs, and PSPs, including Santander, ClearBank, Payoneer and Mashreq Bank, ThetaRay was recognized by CNBC World’s Top Fintech Companies list in 2026.