Cybersecurity Policy Report, Futures industry insiders examine AI’s current benefits, obstacles, (Aug 21, 2026)

By Suzanne Cosgrove

CFTC Innovation Session: Commission looks to forge roadmap for “compute market dominance.”

“As we think about artificial intelligence, we should start by framing it as an innovation supercycle,” said Walt Lukken, chairman of the CFTC’s Innovation Advisory Committee and President and CEO of the FIA. “For decades, the capital markets and the derivatives markets have evolved through gradual innovation, but today the pace of change is breathtaking and approaching exponential,” he added.

Kicking off a discussion of AI at Thursday’s CFTC innovation Advisory Committee meeting, Lukken posed the question: Which AI developments have the greatest impact on CFTC markets and what risks around AI should the Commission be most concerned about?

“It's difficult for markets and regulators to keep pace [and] financial institutions, many around this room today, are deploying generative AI across trading, appliance, surveillance, customer service, and risk management to help manage this change, Lukken said.

Agentic finance grows. The emergence of agentic finance, where autonomous systems execute financial decisions with limited human intervention, represents opportunities, Lukken noted, but also raises questions about accountability, market readiness, customer protections, operational resilience, and market integrity.

“There are some concerns out there about market manipulation from AI agents trading in in the market, and I guess my reaction to that is that algorithmic trading or bots—this is really nothing new, right?” said Brian Armstrong, CEO and co-founder of Coinbase, “We used to call it machine learning. Before that, we called it statistics. Now we call it AI. In many ways, it's just one continuum of the same concept,” he said.

Armstong also predicted stablecoins will be a default payment for AI in the agentic economy. “Increasingly, people are using AI agents to get work done on their behalf, and they're going to need to pay for goods and services.” They're going to need to spin up sub teams of other agents, and the traditional financial system doesn't do as well with very small and real-time cross-border transactions, he said.

Cybersecurity warning. “I think we've seen lots of big cybersecurity issues in the industry,” said Robinhood CEO Vladimir Tenev. Every week, you hear about another protocol sort of getting breached, or some functional correctness issue leading to money being drained, and now you have AI agents that are that are doing extreme offensive cybersecurity work.

Tenev said a solution may be the formal verification of software, “so you can actually prove that the software is correct. And I think it's in in many ways the end game.”

“As the technology gets there, you guys impose more and more stringent requirements for whether code is formally verified and not just … audited by hand,” he said. “The business continuity requirements that you have FCMs conform to can be … tightened up a little bit so that they're relevant in the modern age,” Tenev said.

Facilitating connectivity. As CFTC Chairman Michael Selig noted in his introductory address at the meeting, access to advanced GPU clusters and compute capacity “increasingly determines who can compete, who can innovate, and ultimately, who can lead.”

“As demand for compute grows, so too does the need for markets capable of efficiently allocating scarce resources and managing risk, Selig said.

Weighing in, Tyler Winklevoss, co-founder and CEO of Gemini, said regulators like the CFTC could help by connecting projects with Frontier Labs, foundations or security researchers that want to help harden networks.

For example, a project called Zcash has modeled how to use AI effectively to find vulnerabilities and patch them and then harden the network, Winklevoss said. Zcash is also pioneering formal verification. “But it's a larger project, it's over a decade old, and the smaller projects may not have that connectivity to the Frontier Lab,” Winklevoss said.

“So, to the extent that you can facilitate that connectivity, that could be a great way where regulators could help security researchers get these AI frontier models into the hands of the right people in advance of the adversaries,” he said.

Data centers. The other side of AI is the compute and the data center side, “which is very, very relevant to the CFTC,” said Raghu Yarlagadda, co-founder and CEO of FalconX, a Silicon Valley digital asset brokerage. “Compute probably perhaps is the greatest commodity that ever [was],” he said.

Large U.S. tech companies are expected to invest $670 billion this year in AI infrastructure, with much of that spending going toward data center construction and expansion, he noted, “without knowing the price point or the user demand in two [or] three years.”

“I think there's a remarkable opportunity for CFTC to actually define compute as an asset class,” Yaragadda said. As Frontier Labs are going to public markets for the first time they are getting questions in their IPO in terms of the data center spend and hedging of data center compute, he said.

“At FalconX, what we're seeing is some of these frontier labs in the data centers come and ask, ‘could you write hedging or insurance products around the data center compute?’ he said.

The more the CFTC can guide us in terms of price discovery, fast-tracking regulation around compute, the safer AI will be as an industry in the U.S., and help AI in the U.S. lead worldwide, Yarlagadda said.

Focus on behavior. Thomas Chippas, CEO of the derivatives exchange Rothera, added that “ultimately, the regulator today doesn't know if it came from someone's Excel spreadsheet with a script attached to it, or if someone went on an AI tool and generated a model and it runs in a high-speed fashion.”

“So I think we should consider that it's the behavior that needs to be looked at, maybe less so than the tool. And one of the aspects of behavior is where did this order come from? Where was this quote submitted from?” Consideration of whether or not identification for systematic models might be something to be incorporated, Chippas said.

“It's unlikely that any regulator can keep up with the pace of change that AI is going to generate in terms of new models, new code, new means of execution, but holding models accountable for the same sort of behavioral requirements in markets that we hold everyone accountable for today is a pretty easy principle to agree to,” he said.

RegulatoryActivity: AINews CFTCNews CyberPrivacyFeed DataSecurity ExchangesMarketRegulation DataPrivacy