CFTC wants to transform enforcement work through AI and data ‘marathon’

Few agency workers across the federal government are more closely and publicly associated with their area of expertise than Ted Kaouk is with data. 

A Naval Academy grad with a PhD in English from Maryland, Kaouk carved out that association during lengthy chief data officer stints with the Department of Agriculture and Office of Personnel Management, plus a three-and-a-half-year term as chair of the Federal CDO Council. 

Despite the seen-it-all quality to Kaouk’s federal government career, serving as chief artificial intelligence officer and CDO at the Commodity Futures Trading Commission since December has kicked things up a notch. 

“We ingest and manage over 15 billion records per day. Per day,” Kaouk said of the CFTC, the independent federal agency charged with regulating the U.S. derivatives markets. “So hopefully it gives a sense of the volume of data that we’re managing.”

That data, Kaouk told FedScoop, is what helps the CFTC oversee market participants. And it essentially serves as a foundation for Kaouk’s work within the agency as it undergoes what he calls a “significant” transformation of its oversight and surveillance enforcement capabilities.

The evolution was already underway when Kaouk joined the agency at the end of 2023, part of a set of digitally minded priorities from CFTC Chairman Rostin Behnam.

“He really likens this initiative to a marathon,” Kaouk said of the agency head. “It’s not a sprint, but it’s a continuous effort to ensure our regulatory framework remains … robust and adaptive in an increasingly complex financial landscape.”

The CFTC in recent months has been especially attuned to that landscape in the age of artificial intelligence, issuing a request for comment in January on the uses and risks the technology poses to the derivatives market, which encompasses futures options or contracts for underlying assets such as commodities, bonds, stocks and currencies. The comment period closed in late April and CFTC staffers are currently reviewing submissions, an agency spokesperson said.

The agency followed up its request for comment with the May release of a report on responsible AI in financial markets from a CFTC subcommittee on emerging and evolving technologies. The report’s authors noted that AI has a clear value proposition for the CFTC and its regulatory mission, specifically its ability to automate processes such as surveillance and fraud detection. 

But there remains a “lack of direct knowledge about the CFTC-registered entities currently leveraging AI,” the report noted, underscoring a lack of transparency around everything from the quality of data training to the extent of human involvement in autonomous trading models.

As the agency’s first chief AI officer, Kaouk is pursuing an optimistic approach to the technology and how it can be used to root out fraud manipulation and market abuse.

The CFTC wants to “send a strong signal to bad actors that misconduct will be identified and punished,” he said. “I personally feel that we can harness that data to develop and deploy sophisticated analytics and AI capabilities.”

Compiling the CFTC’s AI use case inventory is in progress and the agency is on track to meet Office of Management and Budget timelines, Kaouk said. In the meantime, he and his team are full steam ahead on a handful of pilots that flex the technology’s data-related muscles. 

For starters, there’s a machine-learning tool that detects anomalies in algorithms that’s focused on improving data quality. There’s another tool that uses AI/ML techniques to identify spoofing behavior in agency data. And there’s a generative AI pilot the agency is in the process of launching that will complement and streamline existing legal research, software development workflows and “general knowledge work,” Kaouk said.

The use case to flag instances of spoofing — the practice in which traders place a bid or offer on a contract or option and then cancel before execution — would be an especially helpful tool for the CFTC’s enforcement arm. 

“The spoofing use case involved using various AI techniques to improve the staff’s ability to identify buy-and-sell offer patterns that may be evidence of potentially disruptive trading patterns that warrant further investigation,” Kaouk said. 

Though the CFTC is an independent federal agency, exempting it from some OMB AI requirements of Chief Financial Officers Act agencies, Kaouk said they have “taken some additional steps” beyond federal guidance and President Joe Biden’s AI executive order, specifically with the establishment of a new section focused on AI and enterprise analytics that sits within the CFTC’s data division. 

“We’re going to be leveraging our existing data governance board to include AI governance,” he added. “So a data and AI governance board, that’s not a requirement. But we feel it’s important for our compliance and our efforts to be able to do that.”

The CFTC also plans to develop an integrated data and AI strategy — another move that is not required of the agency — and is looking to provide training to its workforce in the coming months on AI governance and ethics.  

Having previously led the development of USDA’s first data strategy — which included the creation of an enterprise analytics and AI platform — and guiding OPM through the same process, Kaouk finds himself uniquely well positioned to oversee the CFTC’s data and AI journey. 

There’s still the matter of handling those 15 billion records per day, which comprises everything from swap transactions and price volume figures to whistleblower complaints and consumer tips. But Kaouk believes that lessons learned at USDA and OPM about improving the accessibility and usability of data tools for agency staffers at scale will guide his way.

“At the CFTC, I’m really looking to kind of leverage some of those principles to advance our data and AI capabilities,” he said. “So we’re excited to be moving ahead with exploring how generative AI may be used at the commission in the future.”