Five Minutes with Skadden's Dan Michael

INTERVIEW
On Regulating Increasingly AI-driven markets

As AI reshapes financial services, regulators are grappling with how to oversee the transformative technology. The U.S. Securities and Exchange Commission is tasked with balancing innovation and investor protection in an increasingly AI-driven market.

To understand the SEC's approach, I spoke with Dan Michael, a partner at Skadden in New York City and former head of the complex financial instruments group in the SEC's enforcement division.

While the agency has initially focused on clear-cut cases of misrepresentation, more sophisticated enforcement actions are likely on the horizon. Michael draws parallels between the SEC's AI approach to its stance on cryptocurrency.

The discussion delves into:

• The current status of proposed rules, like predictive analytics 

• The challenges of using AI while complying with existing regulations like Regulation Best Interest

• The complexities of proving intent in AI-driven trading

• The regulatory implications of AI's ability to process vast amounts of unstructured data

We also have some fun imagining the SEC of the future — where an enforcement attorney is tasked with interrogating an algorithm.

At Skadden, Michael is co-head of the law firm’s Web3 and Digital Assets Group, where he represents companies, executives and directors in connection with SEC and FINRA probes and examinations.

Hope you enjoy our chat!

This interview has been edited for clarity and length. 

How is the SEC responding to AI’s growing role in financial advice?

The SEC’s initial approach focuses on “low-hanging fruit,” targeting clear cases of misstatements or fraud, where companies claim their AI can perform functions that it can’t. These types of cases resemble traditional fraud cases but involve a new technology.

As AI’s use in finance evolves, the SEC is likely to pursue more sophisticated cases directly related to the technology’s impact and whether it complies with the federal securities laws. For example, if AI is used to offer financial advice, this raises significant concerns under Regulation Best Interest (Reg BI) and duty of care standards.

What's the status of the SEC's proposed rules on AI and predictive data analytics, and how might they impact the industry?

The SEC has proposed rules related to predictive data analytics, but their status is uncertain. These rules, if implemented, could significantly impact how firms use AI. However, there are concerns about their compatibility with AI technology. The proposed rules may require firms to understand and document every decision made by their AI models, which could be challenging or even impossible with current AI systems.

Chair Gensler seemed to take his foot off the gas in pushing these rules forward when he recently discussed their status in Congress. This could be due to significant pushback from the industry and perhaps influenced by the SEC's recent defeat in the 5th Circuit regarding the private funds rule.

The timing is also notable, as we're approaching an election year. All of these factors combined may have led the SEC to be more cautious about pressing forward with these rules at this time. However, it's important to note that even without new rules, the SEC has considerable flexibility under existing regulations to address AI-related issues. Firms should stay informed about these potential regulatory changes while ensuring they comply with current rules.

How does the SEC view AI in comparison to other emerging technologies, like cryptocurrency?

The SEC’s stance on AI may mirror its approach to cryptocurrency, where it was less inclined to modify existing rules to accommodate new technology. However, unlike crypto, firms using AI are already regulated by the SEC. The conservative mindset of firms already under SEC regulation might slow the adoption of AI, but at the same time may also facilitate a dialogue with the SEC because the firms that want to use the technology have longstanding relationships and ongoing dialogue with their regulator.

What are the primary regulatory concerns when it comes to using AI for financial decision-making?

Key concerns revolve around:

• Regulation Best Interest (Reg BI): Ensuring that financial advice provided by AI tools is in the client’s best interest, with a clear duty of care and suitable recommendations.

• Cybersecurity and Data Privacy: AI systems rely on large datasets, raising questions about data sourcing, storage, and privacy.

• Disclosure Requirements: Companies must accurately disclose how they use AI in their operations. Misstatements or misleading claims about AI’s capabilities could result in enforcement actions similar to those seen in cases involving biotech firms, where the companies made misstatements about a complex drug.

How might AI's capacity to handle unstructured data impact financial analysis?

AI's ability to transform unstructured data (e.g., PDFs, government filings) into structured formats opens up new possibilities for financial analysis. This capability is a game-changer for sophisticated hedge funds, allowing them to use AI tools to analyze vast amounts of data in ways humans never could.

It also raises potential legal and ethical concerns. The App Annie case is a relevant example. App Annie, a data aggregator, collected proprietary data from various sources, claiming it would anonymize and aggregate the information before sharing it. However, the company failed to fully sanitize the material non-public information (MNPI) before sending it to clients.

The SEC charged App Annie with anti-fraud violations, as hedge funds and other clients were using this improperly sanitized MNPI in their trading decisions. This case highlights the importance of understanding data sources and conducting due diligence, especially when dealing with AI systems that can process and utilize vast amounts of data from various sources.

Are there potential risks with AI-powered trading?

Yes, AI trading poses unique risks. For instance, AI models could “learn” to engage in behaviors that might be economically advantageous but illegal, such as insider trading or market manipulation. There’s already evidence from studies showing that AI, under simulated pressure, engaged in actions like insider trading despite being programmed to avoid such conduct.

I came across a study where researchers found that the machines started to collude without explicitly coordinating with each other.

That conclusion doesn't surprise me actually. People “spoof” the market because it's economically advantageous. But it's also illegal. And so I can see situations where the models, just because it is to their economic advantage, engage in trading behavior that is prohibited by the anti-fraud provisions. That's a case that makes me wish I was back in government because how are you going to prove scienter? I'm sure there’s a way you could prove an intent to deceive.

I can’t wait until there’s a movie where an SEC attorney is interrogating a computer. What advice would you give to firms considering the adoption of AI tools?

Firms should ensure that their use of AI complies with existing regulations, especially regarding fiduciary duties and disclosure requirements. They must also be vigilant about data sourcing, cybersecurity, and the transparency of their AI tools’ decision-making processes. Engaging with regulators and staying informed about potential regulatory changes is crucial.

How do you see AI impacting the financial industry over the next few years?

AI will likely continue to revolutionize financial services by enhancing data analysis, improving trading strategies, and automating various functions. However, I think this evolution will be gradual, shaped by regulatory developments and firms’ willingness to navigate the complex compliance landscape. Some firms may take a conservative approach, while others might push boundaries to gain a competitive edge.

Thanks for reading!

Drop me a line if you have story ideas, research, or upcoming conferences to share. [email protected]