AI industry's calls for regulation fail to strike chord with skeptical US antitrust enforcers

The AI industry’s ostensible appetite for regulation is receiving raised eyebrows from top US antitrust enforcers who have pledged to act sooner rather than later to clamp down on anticompetitive conduct such as monopoly maintenance and leveraging of market power by tech companies. The White House has pledged more oversight of companies creating AI foundation models, and US President Joe Biden’s top antitrust officials are keen to not repeat perceived errors that, for instance, allowed Google to turn into a multi-billion dollar Internet search and advertising behemoth that leaves little room for competitors to survive and thrive.

Read on for the story from MLex®, or start your 14-day free trial now to access more specialist news and in-depth analysis in real time.

22 November 2023
By Chris May and Khushita Vasant

The AI industry’s ostensible appetite for regulation is receiving raised eyebrows from top US antitrust enforcers who have pledged to act sooner rather than later to clamp down on anticompetitive conduct such as monopoly maintenance and leveraging of market power by tech companies. 

The White House has pledged more oversight of companies creating AI foundation models, and US President Joe Biden’s top antitrust officials are keen to not repeat perceived errors that, for instance, allowed Google to turn into a multi-billion dollar Internet search and advertising behemoth that leaves little room for competitors to survive and thrive.

“We need to be smarter and more concerned about how algorithms are using data and how companies are setting the objective function,” Doha Mekki, principal deputy assistant attorney general in the Department of Justice's antitrust division, said at an event.*

Mekki said it's "pretty easy" to see that AI products and services are things antitrust enforcers already understand. Rather than imagining AI to be a "brand new paradigm" or set of products, antitrust enforcers need to see it as something that exists within a technological continuum. Privacy regulators have taken a similar stance that existing law applies to generative AI.

“I think monopoly maintenance, leveraging of market power and an existing product seems to be a paradigm that could be very useful if we were to have a concern based on the facts and the law in AI,” Mekki said. “Another paradigm that … we really need to pay a lot of attention to is information-sharing.”

A “very permissive regime” for information sharing revoked by the DOJ in February previously made it possible to maintain collusion, she said.

The DOJ enforcer offered a cautionary tale about anticompetitive conduct in fast-moving digital markets.

Referring to the historic 10-week trial that just wrapped up in the DOJ's suit accusing Google of monopolizing the Internet search and search advertising markets, Mekki said it took the DOJ a "long time to get there."

The agency learned a lot about antitrust enforcement against monopolization and first-wave technologies, which will "make sure that we don't repeat those mistakes when it comes to new horizon or new frontier AI technologies.”

"Companies were very good at saying, ‘Don't use antitrust, that's the sledgehammer. Just regulate us,’” Mekki said .

Inputs that feed generative AI products include massive data caches that have already been indexed, Mekki said. So, it stands to reason that a company that already has an advantage in Internet search is probably going to “have a leg up” in certain kinds of AI technologies.

The DOJ enforcer was apparently hinting at Google, where multiple witnesses testified that user data and scale that derived from user data was critical for the tech giant to be able to pull in billions in revenue.

“If you have a leg up in cloud computing, you probably are going to have a leg up, again, in new horizon AI technologies," Mekki said. “You know, I worry sometimes that clever marketing and PR sort of make these products seem new, but they are not, in fact, new.”

“One thing that's become clear is an algorithm is just software like the code itself is not all that useful. What is very useful is data, particularly the proprietary time and the objective function,” she said.

“As deep learning advances, one can imagine that these tools will get a lot better at efficiently colluding in a way that humans can't,” Mekki said.

Regulation, legislation, misinformation

The poster child for the AI industry’s pro-regulation charm offensive is OpenAI's controversial frontman, Sam Altman, who pursued collaboration with legislators and regulators across the globe in an effort to promote responsible innovation while averting a robot-fueled apocalypse.

After a dispute with OpenAI’s board that had Altman and much of his team headed to Microsoft over the weekend, OpenAI said in a Tweet late Tuesday that Altman would return as CEO.

In response to a question asking Mekki to call out “bits of disinformation and misinformation” around the AI industry, she encouraged more scrutiny of claims from companies that either are already offering AI services or stand to benefit from the explosion of AI, “and particularly how they think about regulation and antitrust.”

Record-breaking lobbying outlays from Big Tech in 2022 that topped $70 million and helped derail bipartisan antitrust legislation before Congress gave Mekki pause, she said. “The argument seems to be we want regulation, just not that regulation. And I think we should be careful about that.”

California's privacy agency is expected to discuss broad new regulations on machine-learning technology at a meeting December 8, while the regulation of "foundation" or "general purpose" AI models under the European Commission's proposed AI Act remains a key sticking point for EU lawmakers.

US Federal Trade Commission Chair Lina Khan recently channeled concerns from AI start-ups that a series of confidential confabs between Big Tech and Congress could prove fertile ground for pro-monopoly regulatory capture.

Katie McInnis, the top Democratic counsel on the House Judiciary Committee's antitrust subcommittee, called AI company requests for regulation “highly clever.”

“They learned from years and years of hearings with Big Tech executives that pushing back on regulation isn't actually helpful there,” McInnis said at a recent event that included a representative from OpenAI. **

“You can say, ‘We welcome it,’ and then we're all going to scramble around for a while trying to figure out what that means … and you've sufficiently diverted attention away from some of the problems here,” McInnis said.

OpenAI’s in-house competition counsel, Haidee Schwartz, who spoke at the same event as McInnis, said the industry’s desire to be regulated is not to “divert” from existing laws governing AI, but rather to address “where there is existential risk for the most frontier AI systems.”

“You would not want to stifle new entrants or new development,” she said, and by using a tiered licensing regime, “you would hope that that would work.”

Departures of high-profile and bipartisan proponents of antitrust action in McInnis’ committee underline a point that hardly needs arguing these days: Passing new legislation through Congress anytime soon — antitrust or otherwise — will be a rough road.

Bringing the hammer down

While the White House moves forward with executing a privacy-focused executive order that expands on voluntary agreements brokered with some of the biggest companies working in the AI space, Biden administration antitrust enforcers aren’t waiting around to make their own moves.

Mekki, who supervises more than 350 attorneys investigating and prosecuting civil and criminal violations of federal antitrust and other competition laws, has also spent a lot of time thinking about information sharing this year, she said.

The DOJ recently backed private price-fixing litigation against software company RealPage and rental property managers who are accused of “knowingly combining their sensitive, nonpublic pricing and supply information in an algorithm that they rely upon in making pricing decisions, with the knowledge and expectation that other competitors will do the same.”

Another senior DOJ official linked AI-fueled information sharing to inflated prices for kitchen staples in the context of a recent lawsuit against Agri Stats. The DOJ and four states accuse the data broker of encouraging US meatpackers to inflate profits through an illegal, multi-industry scheme to distribute competitively sensitive information through exclusive subscription and consulting services. This reinforces DOJ antitrust chief Jonathan Kanter’s focus on the meat packing and processing industry, a “kitchen table issue” that the agency is taking seriously.

Meanwhile, Khan’s FTC filed a watershed suit against Amazon, taking aim at its multiple algorithmic pricing systems under revamped competition authority outlined in Section 5 of the FTC Act.

The DOJ and FTC are battle-ready to take on the antitrust challenges posed by AI by ramping up the recruitment of experts in computer programming and analysts for extremely large datasets.

—With additional reporting by Serafina Smith

*AI and the Public Interest, organized by the Open Markets Institute, Washington, DC, Nov. 15, 2023.

**"2023 Fall Forum: Can Antitrust and Consumer Protection Keep Up with Artificial Intelligence? What You Need to Know," American Bar Association; Washington, DC and online; Nov. 9, 2023.

For access to breaking news and predictive analysis on regulatory risk across the globe in real time, start your free trial today.