The SEC’s misguided attempt to regulate AI

My colleague John Berlau and I submitted a comment letter this week to the Securities and Exchange Commission (SEC), arguing its proposed regulations on artificial intelligence and predictive data analytics are fundamentally flawed. While perhaps well-intentioned, the agency’s sweeping new proposal epitomizes the very “behavioral biases” it aims to restrict.

The proposed regulation requires investment advisers and broker dealers to identify and then eliminate conflicts of interests surrounding the use of a broad range of covered technologies, in addition to creating new record-keeping requirements. However, the benefits of the regulation for investors are unclear. The SEC does not estimate any benefits or fully analyze a suite of alternatives.

In short, we don’t believe the SEC has made a convincing case that a systemic problem exists in the innovative marketplace for investor trading. Beyond a few anecdotes, the SEC provides no rigorous empirical evidence of concrete harms to investors from these technologies. The proposal instead relies on speculation, rather than data.

In all likelihood what has happened is that, in its haste to react to GameStop’s stock price skyrocketing in early 2021, the SEC has skipped key analytical steps. Seeking to appease public outcry and respond to a perceived urgency to “do something,” the SEC did not adequately weigh costs, benefits and alternatives before rushing to propose new rules.

This rushed process produced a proposal the SEC estimates will impose $460 million in upfront and $230 million in ongoing compliance costs, all without any quantified benefits. The SEC failed to fully analyze alternatives like standardized disclosure that may achieve its goals more efficiently. This incomplete analysis offers little reason for confidence the chosen proposal will protect investors.

Essentially, the SEC is committing the very same sin it accuses consumers of electronic trading apps like Robinhood of—letting biases and short-sighted thinking cloud sound judgment. The difference is, flawed regulations impose collateral damage across the economy, while individual investors bear the costs of their own poor investing decisions.

Before imposing sweeping new rules, the SEC needs far more evidence that a concrete problem exists, that its proposed solution will directly address that problem, and that the alternative selected will solve the problem in a cost-efficient manner. Absent that foundation, the regulation will likely introduce new inefficiencies and biases into capital markets.

In reality, AI and predictive data analytics offer tremendous potential to improve decision-making and expand access to investment advice, all while making trading easy and fun. Rather than prematurely restrict new technologies, regulators should instead adopt a more permissionless approach that allows experimentation and discovery of best practices. The SEC would be wise to withdraw this proposal and pursue a more innovation-friendly path.