Although enactment of federal legislation regulating the use of AI is likely not on the near-term horizon, a number of federal agencies, including the Federal Trade Commission (FTC), have been reminding the entities they regulate that use of AI must still comply with current laws and regulations. In the case of the FTC, this has meant ensuring that companies under its purview do not use AI to commit unfair or deceptive acts under Section 5 of the FTC Act.
In furtherance of that policy goal, on August 16, 2023, a staff attorney at the FTC Division of Advertising Practices published a business blog post on the FTC’s website entitled “Can’t Lose What You Never Had: Claims About Digital Ownership and Creation in the Age of Generative AI.” In the post, the FTC noted that improvements in the ease of use and quality of generative AI are enabling digital works generated by AI tools to be increasingly passed off as the work of human authors. The FTC outlined four key points companies should keep in mind when working with digital products, in some cases whether or not AI-related, in order to avoid misleading consumers or other businesses:
- - When offering a generative AI product, companies may need to inform customers whether and the extent to which AI training data includes copyrighted or otherwise protected material.
- - Companies should not try to “fool people” into thinking that AI-generated works were created by humans.
- - Companies must ensure that customers understand the material terms and conditions associated with digital products, since those terms often differ from terms associated with nondigital goods (e.g., digital goods are often licensed and not sold). The FTC also noted that unilaterally changing terms or undermining reasonable ownership expectations can be problematic.
- - Companies that offer a platform for creators to develop and display their work must be clear about the creators’ right to access and remove their work, as well as how that work will be used and presented.
Of these admonitions, the first one — relating to disclosures about AI training data — is perhaps the most interesting. The FTC is undoubtedly aware that use of copyrighted and “otherwise protected” material in training data without permission has been controversial: The practice sparked a series of lawsuits to determine whether such activity is fair use under copyright law or violates privacy rights, and groups such as the Author’s Guild have published strong statements criticizing the practice. It is noteworthy that, in contrast to the other warnings, the FTC uses the term “may need” with respect to the training data disclosure issue, providing companies with some leeway regarding how to apply this guidance. Nonetheless, the FTC cautions that failing to disclose that copyrighted or otherwise protected materials were used to train an AI tool could raise issues of consumer deception or unfairness, especially where the AI outputs may reflect the use of such material. Although the FTC does not explain what it means by “otherwise protected material,” this could be a reference to personal data, a topic of focus for the FTC in the AI space. The FTC notes that transparency on this point about training sources could be relevant to individuals when they decide which AI tool to use, and could be relevant to businesses since they could face liability if the output they generate through use of an AI tool infringes protected works. Companies offering AI models and tools, many of whom have not disclosed the source of their training data, will need to consider the FTC’s statement on this point.
Ongoing FTC Activity in the AI Space
The FTC’s August post is the latest in a series of statements the FTC has made and actions it has taken against companies regarding AI usage. For example, a March 2023 FTC blog post, “Chatbots, Deepfakes, and Voice Clones: AI Deception for Sale,” cautioned companies about offering AI tools that might be used for deception. The FTC suggested in that post that companies offering AI products consider at the design stage the ways the product might be misused for fraud or to cause harm, and consider whether the product should therefore not be offered. The FTC has also deployed the remedy of “algorithmic disgorgement” in some of the actions the agency has brought, which requires companies found to have collected personal data illegally to delete that data as well as any algorithms trained on that data.
Companies operating in the AI space should monitor the FTC’s evolving guidance on the use of AI, and ensure their practices conform with such guidance.