Artificial Intelligence Pushes the Antitrust Envelope

Keep up with the latest developments and legal issues in the telecommunications and emerging technology sectors, with exclusive access to a comprehensive collection of telecommunications law news,...

By Michaela Ross

If machines collude when no one’s watching, would any antitrust alarm bells sound?

Artificial intelligence is raising concerns in antitrust circles, particularly in the European Union. It’s not clear whether the tools regulators use to gauge competition are sophisticated enough to assess the impact of these new technologies. Elements of AI, such as the algorithms used to make decisions and the big data sets that fuel those decisions, may test the ability of antitrust regulators to assess new forms of collusion and barriers to market entry, antitrust lawyers, academics and tech startup execs say.

In some cases, antitrust authorities will need to adapt current analysis tools to new tech business models, but in other situations, authorities may be left with virtually no avenues to tackle competitive harms, Maurice Stucke, a University of Tennessee antitrust law professor and author on the subject, told Bloomberg BNA.

“There are some anti-competitive risks for big data and big analytics that our current law doesn’t reach,” Stucke said.

Antitrust regulators may be confronted with machines artificially inflating prices across a market unbeknownst to their operators, or tech giants holding massive customer data sets that could give them an unfair advantage in a market. The challenges of big data and algorithms have arisen in recent antitrust cases involving Microsoft Corp.'s merger with LinkedIn Corp., as well as sellers on Inc.

The stakes are only going to get bigger as the pace of both AI development and tech mergers accelerate. Mergers and acquisitions across the tech, telecom and communications sectors have increased almost 40 percent in the last five years, according to Bloomberg Law Deal Analytics data. Tech companies are expected to be swept up in a rising wave of mergers and acquisitions in 2017. More than 200 private companies leveraging artificial intelligence have been acquired since 2012, with more than 30 acquisitions in the first quarter of 2017 alone, according to CB Insights, an investment research company. Actions to regulate algorithms and big data are likely to diverge as EU antitrust authorities scrutinize digital platforms more heavily than their U.S. counterparts.

Data’s Value Conundrum

Many of the acquisitions involve companies such as Alphabet Inc.’s Google, Microsoft Corp. and Facebook Inc. snatching up AI startups. The tech giants are looking for innovative AI algorithms to leverage the large data sets about user behavior they’ve collected into new products and services.

The big data sets that artificial intelligence relies on to learn and operate are difficult to value in a merger or when analyzing individual company conduct, said Jérôme Philippe, partner at Freshfields Bruckhaus Deringer LLP’s Paris office. There is increasing concern, especially in Europe, that the data sets owned by tech giants are so essential to dominating a market that they can act as a barrier to entry to new startups, the firm said in a report on antitrust enforcement.

For example, Microsoft’s acquisition of LinkedIn faced scrutiny from EU regulators who feared LinkedIn’s user data could give the new company an unfair advantage against competitors. The company plans to combine LinkedIn’s data with artificial intelligence algorithms in new sales-focused software. Microsoft assuaged regulators’ concerns and avoided a formal investigation with a string of concessions, such as adding a LinkedIn competitor to its Windows platform. Other data-rich companies may find themselves making similar concessions in a deal review, but it will likely depend on the perceived value of the information set, which can be difficult to determine, antitrust lawyers and trade groups told Bloomberg BNA.

Apart from mergers, the methods U.S. antitrust regulators use to assess potential monopolies are not designed to scrutinize large data sets, Herbert Hovenkamp, antitrust law professor at the University of Iowa, told Bloomberg BNA.

To be sure, antitrust lawyers in the U.S. and EU are quick to note that data accumulation is of itself not an antitrust red flag. Data sets should be treated on a case-by-case basis depending in part on the uniqueness of the data and its scope, they say. Quantifying the value of data or labeling a data set as essential to market entry ignores the fact that data can be shared with multiple companies and can be easily replicable.

“If all we’re talking about is the simple accumulation of data by a firm, like Google, no matter how big, we don’t have any reason to go after them now under antitrust laws,” Hovenkamp said.

For example if an internet user gives his or her age to a social network, that doesn’t prevent other social networks from also obtaining that information, Philippe said. Traditional brick-and-mortar companies such as retailers and insurers have harbored large user data sets for years, but have avoided scrutiny because they haven’t exploited the data in the same way as digital platforms, he said.

“What I’m afraid of is in the end we sanction these companies just because they’re better at using the data,” he said.

Colluding Computers

Another way artificial intelligence systems could fly under the radar of an anti-competitive analysis is through tacit collusion. Price-fixing algorithms at competing companies could coordinate to keep prices artificially high, for example, without any human influence.

“Finding ways to prevent collusion between self-learning algorithms might be one of the biggest challenges that competition law enforcers have ever faced,” the Organization for Economic Co-operation and Development said in an October report. “Particularly in the case of artificial intelligence, there is no legal basis to attribute liability to a computer engineer for having programmed a machine that eventually ‘self-learned’ to co-ordinate prices with other machines,” the report said.

In other words, algorithms programmed to maximize profit may quickly self-learn that the best way to do that is to coordinate with other pricing algorithms to keep prices stable and above market value. There would be no advantage for an algorithm to cut prices in an attempt to attract consumers, as other algorithms would immediately adjust.

Simple pricing algorithms are already changing the competitive dynamic. In March, a third-party seller of posters on Amazon faced criminal antitrust charges in the U.S. District Court of the Northern District of California after pleading guilty to recruiting other sellers to knowingly use a set of price-setting algorithms as part of a cartel scheme. In that case, humans colluded to control algorithms, but the activity may not always be as clear cut. In 2011, two third-party sellers on used competing algorithms that adjusted the price of a textbook according to the other’s price, eventually raising it to $23 million. If a price hike is not the result of an agreement between sellers or a cartel, it is not illegal in the eyes of antitrust authorities, Philippe said.

Antitrust May Diverge Across the Atlantic

Anti-competitive analysis of companies utilizing artificial intelligence technologies has potential to diverge between the U.S. and the EU in the coming years, lawyers and trade associations say, as EU antitrust authorities scrutinize the use of personal data by tech giants. For example, Microsoft’s acquisition of LinkedIn sailed through approvals by U.S. antitrust regulators before facing roadblocks from European authorities.

“A lot of people are afraid of those big companies coming from the U.S. and taking their data, and I think it’s a trend of competition authorities to follow that fear from the people,” Philippe told Bloomberg BNA. “I’m afraid Europe will take the lead in going against big data.”

This is in part spurred by the EU’s model of using antitrust tools to police alleged privacy infringements on digital platforms. Margrethe Vestager, the EU’s commissioner for competition, is investigating Facebook Inc. after it rolled back promises not to share user data between services after acquiring WhatsApp messenger in 2014. German competition regulators last year opened an investigation against Facebook for allegedly using its dominant social platform to undermine user privacy protections.

Vestager is also conducting ongoing investigations against Alphabet Inc.'s Google, alleging the company has programmed its search algorithms—an early example of artificial intelligence technology —to favor its own advertising and shopping service over competitors.

Meanwhile, a legislative proposal in Germany and an initiative at the EU commission are looking into expanding regulators’ abilities to scrutinize mergers when the target may have low income but a high value. These efforts are the result of European concerns that the value of data sets held by WhatsApp during its Facebook merger contributed to the messaging service’s $18 billion value, according to Bloomberg data, but did not trigger an antitrust review notification in the country.

The U.K. House of Lords, French, German and Dutch antitrust authorities as well as the EU Commission and European Data Protection Supervisor have also studied digital platforms, big data and competition law in recent months, with several of their reports surfacing concerns about algorithmic collusion.

Still, Vestager said at an American Bar Association event in late March that she’s not looking for a “war” on algorithms. The antitrust czar earlier said competition enforcers shouldn’t be suspicious of every automated pricing software system or every company with a valuable data set.

French and German antitrust authorities, in their May 2016 analysis of big data, argued that the complex nature of data sets warrant a case-by-case approach, a sentiment echoed by Democratic U.S. Federal Trade Commissioner Terrell McSweeny in public speeches last year.

In the U.S., President Donald Trump hasn’t put in place a full set of antitrust authorities at the FTC and Justice Department. Makan Delrahim, Trump’s pick to lead the department’s antitrust division, has yet to testify before the Senate Judiciary Committee.

FTC commissioners from both parties have maintained that privacy concerns are often best addressed through consumer protection and not competition laws, as outlined by acting FTC Chairwoman Maureen Ohlhausen’s 2015 legal article on the subject.

Under Ohlhausen’s watch, lawyers and academics are not expecting U.S. antitrust authorities to significantly expand competition law.

Artificial intelligence is still in its nascent stages as a technology, and competition authorities on both sides of the Atlantic have said it will take time to evaluate how big data and algorithms may impact antitrust regulation.

To contact the reporter on this story: Michaela Ross in Washington at

To contact the editor responsible for this story: Keith Perine at

Copyright © 2017 The Bureau of National Affairs, Inc. All Rights Reserved.

Request Tech & Telecom on Bloomberg Law