A few key principles: An excerpt from Chair Khan’s Remarks at the January Tech Summit on AI

At the FTC’s Tech Summit on Artificial Intelligence, Chair Khan previewed how the FTC is using its experience and expertise to establish rules of the road for AI. In her remarks, she noted that “as we continue this work, a few key principles come to mind.  

"First, we are focused on scrutinizing any existing or emerging bottlenecks across the AI stack. History shows that firms that capture control over key inputs or distribution channels can use their power to exploit those bottlenecks, extort customers, and maintain their monopolies. The role of antitrust is to guard against bottlenecks achieved through illegal tactics and ensure dominant firms aren’t unlawfully abusing their monopoly power to block innovation and competition.

"The agency is taking a close look across the AI stack to understand the extent of competition across the various layers and sub-layers. We are examining whether dominant firms with control over key inputs—like cloud infrastructure and access to GPUs—may be able to impose coercive terms, charge extractive fees, or deepen their existing moats. We’re keen to sharpen our thinking on the various opportunities and potential obstacles for competition across AI markets and are eager to be learning from players within this ecosystem

"Second, we are squarely focused on how business models drive incentives. Just as we’ve seen behavioral advertising fuel the endless collection of user data, model training is emerging as another feature that could further incentivize surveillance. The FTC’s work has made clear that these business incentives cannot justify violations of the law. The drive to refine your algorithm cannot come at the expense of people’s privacy^[1] or security,^[2] and privileged access to customers’ data cannot be used to undermine competition.^[3] We similarly recognize the ways that consumer protection and competition enforcement are deeply connected—with firms engaging in privacy violations to build market power and the aggregation of market power, in turn, enabling firms to violate consumer protection laws. And our remedies will continue requiring that firms delete models trained on unlawfully acquired data in additional to the data itself.^[4]

"Third, we are squarely focused on aligning liability with capability and control. This requires looking upstream and across layers of the AI stack to pinpoint which actor is driving or enabling the lawbreaking. Our enforcement experience in other domains will directly inform how we approach this work. For example, our recent robocall enforcement sweep not only targeted telemarketers and the companies that hire them, but also looked upstream to the lead generators and Voice over Internet Protocol (VoIP) providers that enable illegal telemarketing.^[5] And in our recent work to combat scams, we are holding upstream payment actors accountable for knowingly facilitating fraud.^[6] 

"And fourth, we’re focused on crafting effective remedies that establish bright-line rules on the development, use, and management of AI inputs. The FTC is making clear that some data is simply off the table for training models. For example, our recent order against Rite Aid bans the company from using facial recognition tools after its reckless application of the technology led to innocent people being accused of shoplifting.^[7] And our recent cases against data brokers include bans on using or monetizing people’s highly sensitive location data.^[8] 

"As we continue to establish rules of the road for AI, it’s essential that we set clear boundaries on the content that can and cannot be used for scraping and model-training. The Commission recently held a public workshop with creative professionals to better understand the types of guardrails that would help protect against creators’ work being appropriated and devalued by generative AI models, including in ways that may undermine fair competition. Our subsequent report on generative AI and the creative economy lays out our core concerns and how our authorities may apply in this space.^[9]

"Across all of our work, we are making clear that there is no AI exemption from the laws on the books.^[10] Firms can’t use claims of innovation as a cover for lawbreaking. And we’ve already made that clear in a number of AI-specific contexts. Model-as-a-service companies that deceive users about how their data is collected may be violating the law.^[11]  Companies claiming that interoperability must come at the expense of privacy and security may be violating the law.^[12] And companies that deceptively change their terms of service to their own advantage may be violating the law.^[13]"

[1] Press Release, Fed. Trade Comm’n, FTC Sues Kochava for Selling Data that Tracks People at Reproductive Health Clinics, Places of Worship, and Other Sensitive Locations (Aug. 29, 2022), https://www.ftc.gov/news-events/news/press-releases/2022/08/ftc-sues-kochava-selling-data-tracks-people-reproductive-health-clinics-places-worship-other.