In April 2018, a PSA featuring former president Barack Obama made its rounds on the internet. In the video, Obama warned that we were “entering an era in which our enemies can make it look like anyone is saying anything at any point in time.” To hammer that point home, Obama referred to then-president Trump as “a total and complete dipsh!t.” Or did he?
The video was actually a deepfake in which actor-director Jordan Peele voiced Obama and an early-version of algorithmic machine learning technology – what we today refer to as “Generative AI” – made it look like Obama was speaking Peele’s words.
It’s been six years since that video launched, and Generative AI technology has evolved to a point where anyone with an app on their phone can create a convincing deepfake. And the proliferation of this technology has many concerned – including the FTC.
Earlier this month, the Commission issued a Supplemental Notice of Proposed Rulemaking (SNPRM) that, if enacted, would extend the reach of the Government and Business Impersonation Rule (16 CFR § 461; “Impersonation Rule”) to (i) prohibit the impersonation of any individual and (ii) impose liability on the entities that provide the tools – namely Generative AI – that allow for such impersonations.
These amendments would not only significantly expand the FTC’s jurisdiction and enforcement efforts over AI but could also lead to increased liability for platforms and providers of Generative AI services.
Let’s break it down.
No Impersonations. Period.
The current Impersonation Rule prohibits the impersonation (or misrepresentation of endorsement) of government and business entities and their respective officers. The first proposed amendment would extend these prohibitions to cover the impersonation (or misrepresentation of endorsement) of any individual – whether real or fictitious.
This would allow the FTC to address harms caused by the recent proliferation of “pig butchering” – where grifters pose as individuals to extract money or sensitive information from consumers – and other relationship-based scams.
Platform Accountability
The second proposed amendment would impose direct liability on entities and individuals who provide goods or services with knowledge or reason to know that the goods or services will be used for such prohibited impersonations (or misrepresentations of endorsement).
This would allow the FTC to hold accountable and seek monetary relief from the platforms and companies that design and offer the Generative AI tools that create images, audio, videos, and/or text that may be used in such prohibited impersonations.
Show Me the Money
The SNPRM is the latest effort from the Commission to regulate AI under its consumer protection authority. It is designed not only to deter bad actors from engaging in this behavior but also to impose greater and more immediate financial liability – including civil penalties for violations – on entities that provide the technology and tools to these bad actors. It further evidences the Commission’s focus on obtaining monetary relief in a post-AMG environment.
Section 18 of the FTC Act (15 USC § 57a) authorizes the Commission to prescribe rules defining as unfair or deceptive certain acts or practices in or affecting commerce. If any person or corporation violates a rule promulgated under Section 18, Section 19 (15 USC § 57b) authorizes the FTC to commence a civil action in federal court and to seek monetary relief in the form of disgorgement and damages.
Thus, by declaring AI impersonation an inherently “unfair or deceptive act or practice,” the FTC will have the ability to directly file a case in federal court and seek monetary relief – in addition to injunctive relief – for any rule violation.
Comments?
The public commenting period for the SNPRM will be open for 60 days following publication in the Federal Register, which should occur in the next few weeks.
Stay tuned to this space for more updates.
[View source.]
