On September 19, 2024, the Institute for Free Speech submitted comments to the Federal Communications Commission to express opposition to the proposed rulemaking “Disclosure and Transparency of Artificial Intelligence-Generated Content in Political Advertisements, Docket No. 24-211.”

Read a PDF of the comments here.

September 19, 2024

Federal Communications Commission
45 L Street, N.E.
Washington, D.C. 20554

RE: Comments on Notice of Proposed Rulemaking: Disclosure and Transparency of Artificial Intelligence-Generated Content in Political Advertisements, Docket No. 24-211

Dear Commissioners,

On behalf of the Institute for Free Speech,[1] I am writing to express our opposition to the proposed rulemaking “Disclosure and Transparency of Artificial Intelligence-Generated Content in Political Advertisements, Docket No. 24-211” (“AI Rulemaking” or the “Proposed Rule”).

I. Summary

While there are many problems with this proposal, a few stand above the others and are fatal to this effort.

First, the Commission lacks statutory authority to require disclosure or disclaimers on political advertisements that utilize “artificial intelligence.” Congress knows how to require disclaimers on political advertisements. Indeed, it has done so repeatedly, explicitly requiring certain political communications to include “paid for by” disclaimers and “stand by your ad” statements. It has adopted no analogous requirement for AI-generated content.

In the absence of Congressional authorization, the FCC cannot rely on its general authority to adopt rules as “necessary” to carry out the provision of the Federal Communications Act of 1934, as amended (the “Act”). This is not a free-floating grant of authority. It requires a tie to another provision of the Act. And there is no provision concerning AI-generated content.

There is no general authority for the FCC to police the truth or falsity of political advertisements, nor could there be. This is particularly true for candidate broadcast advertising, where the Act itself explicitly denies broadcasters the ability—let alone the duty—to remove ads for any reason, including the truth or falsity of the communication.[2]

Second, even if the Commission could regulate AI-generated content in political advertisements—and it cannot—key definitions are overly broad and impermissibly vague.

Proponents of the Rule point to the risk of so-called “deepfakes,” made possible by advances in machine-learning capabilities. But the definition of AI-generated content proposed by the Commission is untethered from these technological advances. Instead, it refers to content “generated using computational technology.” This would seem to encompass any content created in whole or in part using computers, which is to say, the vast majority of advertising content. Assuming the Commission does not intend for the Proposed Rule to sweep so broadly, people of ordinary intelligence must necessarily guess at what might be covered. This guesswork is inherently chilling and is not permitted under the First Amendment.

In addition, rather than relying on defined categories of political communications—such as express advocacy or electioneering communications, the Commission proposes adopting its own idiosyncratic definition of “political advertising.” In doing so, the Commission relies on an undefined term, “political matter,” and proposes an inherently subjective definition of “issue advocacy” that fails to provide proper notice of what is regulated.

Next, the Commission’s cavalier approach toward political speech is reflected in its proposed regulatory flexibility analysis. Here, the Commission disregards the impact of the proposed rule on candidate campaigns, political committees, and advocacy groups—i.e., the very entities whose First Amendment speech rights would be diminished by this regulatory overreach. Instead, the Commission focuses only on broadcasting companies and carriers as small entities that may be impacted. This myopic approach is contrary to law and further highlights that the FCC is the wrong agency to be regulating political speech.

Finally, the proposed rule poses several practical challenges, even on its own terms. For example, when the Proposed Rule is combined with other required disclaimers, it begins to occupy a significant portion of a radio advertisement, limiting the ability of speakers to convey their own messages.

Reasonable people may disagree about what, if anything, should be done about AI-generated content in political advertisements. But those decisions are for Congress to consider in the first instance. Not unelected officials in an agency whose core mission is not regulating political speech.

This Commission can and should reject this Rule and allow the proper democratic processes to play out.

I. The Commission Lacks Authority to Require Additional Disclaimers or Disclosures for Political Communications

The question for the FCC “is not whether something should be done; it is who has the authority to do it.”[3] “Agencies have only those powers given to them by Congress, and ‘enabling legislation’ is generally not an ‘open book to which the agency [may] add pages and change the plot line.’”[4] Yet, that is precisely what the Commission seeks to do with this Rule.

a. The Negative Inference of Congressional Adoption of Some Disclaimer Requirements is that There is No Free-Floating Authority to Create Others

Consistent with the basic cannon of statutory construction expressio unius est exclusion alterius, the adoption of some disclaimer requirements implies the exclusion of authority for others.[5]

Congress knows how to require disclaimers on political communications. Indeed, it has done so—explicitly—as part of the Federal Election Campaign Act of 1971, as amended (“FECA”). Under FECA, “political committees” and others who make independent expenditures or “electioneering communications” are required to include specific disclaimers stating who paid for the communication, whether it is authorized by a candidate or campaign committee, and, for communications that are not authorized by a committee, provide a street address, phone number, or website address for the speaker.[6] As part of the Bipartisan Campaign Finance Reform Act (“BCRA”), Congress added “stand by your ad” requirements, mandating that candidates and other persons making independent expenditures by radio or television include a statement by the candidate or a representative of a non-candidate speaker stating that they approved the communication or are responsible for the content of the message.[7]

There is no analogous requirement for AI-generated content in the Act. On the contrary, the Act includes specific requirements for what must be included in the “political record”:

• whether the request to purchase broadcast time is accepted or rejected by the licensee;
• the rate charged for the broadcast time;
• the date and time on which the communication is aired;
• the class of time that is purchased;
• the name of the candidate to which the communication refers and the office to which the candidate is seeking election, the election to which the communication refers, or the issue to which the communication refers (as applicable);
• in the case of a request made by, or on behalf of, a candidate, the name of the candidate, the authorized committee of the candidate, and the treasurer of such committee; and
• in the case of any other request, the name of the person purchasing the time, the name, address, and phone number of a contact person for such person, and a list of the chief executive officers or members of the executive committee or of the board of directors of such person.[8]

It makes no mention of AI or AI-generated content.

The fact that Congress has adopted some disclaimer requirements necessarily excludes other, heretofore unknown requirements. The negative inference is even stronger with respect to the inclusion of information regarding AI-generated communications in political file disclosures.

Importantly, Congress has entrusted the enforcement of this statutory scheme not to the FCC, but to the Federal Election Commission.[9] This regulatory scheme is so comprehensive that the D.C. Circuit—in an opinion written by then-Judge Ruth Bader Ginsburg—held, “the FEC is the exclusive administrative arbiter of questions concerning the name identifications and disclaimers of organizations soliciting political contributions.”[10]

If Congress wished for additional disclaimers and disclosure on political ads, or wished to delegate such authority to the FCC either in place of or in addition to the FEC, it knows how to do so. It has done neither, and the Commission lacks the authority to do so now.

b. Section 303(r) Does Not Provide Authority for the Rule

The Supreme Court has cautioned that “‘enabling legislation’ is generally not an ‘open book to which the agency [may] add pages and change the plot line.’”[11] Recognizing that the FCC has no specific authority for the Rule, the Notice of Proposed Rulemaking falls back on the general provisions of section 303(r). This is little more than a regulatory power grab that is ultimately unavailing.

Section 303(r) provides the Commission authority to “[m]ake such rules and regulations and prescribe such restrictions and conditions, not inconsistent with law, as may be necessary to carry out the provisions of this chapter, or any international radio or wire communications treaty or convention, or regulations annexed thereto, including any treaty or convention insofar as it relates to the use of radio, to which the United States is or may hereafter become a party.”[12]

By its own terms, the rulemaking authority in section 303(r) must be tied to other provisions of the Act or specific treaties or conventions entered into by the United States. But there is no other provision of the Act to which the Rule relates back. Thus, the Commission’s citation to section 303(r) is circular: section 303(r) only provides authority to adopt rules to implement other provisions of the Act. The predicate cannot be section 303(r) itself.

In response, the Commission points to the “public interest” licensing criteria of sections 307(a) and 309(a). But these citations are inapposite.

The purported “public interest” advanced is “address[ing] false, misleading, or deceptive material and [] ensur[ing] that voters have the information needed to assess the reliability and credibility of political ads.” But policing the content of political advertising for accuracy is not an “interest” within the FCC’s purview. Indeed, if anything, the statutes passed by Congress make clear that the FCC and its licensees are not supposed to act as arbiters of truth in the political arena. To wit, the Act specifically provides “[t]hat such licensee shall have no power of censorship over” candidate advertisements.[13]

II. The Proposed Rule Would Impermissibly Chill Political Speech Through Vague and Overbroad Definitions 

Even if the Commission had authority to regulate in this area, this rule is substantially flawed and should be rejected. 

a. The Proposed Definition of AI-Generated Content is Overbroad and Impermissibly Vague

The Supreme Court has warned, “Prolix laws chill speech for the same reason that vague laws chill speech: People ‘of common intelligence must necessarily guess at [the ’law’s] meaning and differ as to its application.’”[14] The proposed definition of “artificial intelligence-generated content” suffers from the same infirmity—people of ordinary intelligence must necessarily guess what it means—and risks the same adverse consequences—chilling political speech.

As written, the proposed rule seeks to regulate images, audio, or video “generated using computational technology or other machine-based system that depicts an individual’s appearance, speech, conduct, or an event, circumstances, or situation.” The natural reading of this definition would include any communication created with or edited in part on a computer or other similar technology.

Given that the use of cheap, commonly available editing software and tools is ubiquitous in the creation of political communications, this definition could easily encompass every political advertisement. Including an AI-generated advertisement disclaimer on every advertisement plainly is not what the Commission intends. Indeed, doing so would defeat the express purpose of the proposed rule: to identify those advertisements using new and potentially misleading technology.

As a result, it is reasonable to conclude that the Commission would not read this definition quite so literally. The problem is that once the Commission departs from the plain text of its own regulation, speakers and broadcasters have no principled basis for determining when a disclaimer is required. For example:

• How would the rule apply to AI-generated advertisements that are unambiguously intended to be satire? Speakers have long used computer-generated content to parody political figures[15]—would all of those communications now require disclaimers?
• What about real images that are altered using common photo editing software? Recent upgrades to such software often incorporate AI to improve results.[16] Campaigns have long used color saturation to depict candidates in a positive or negative light, such as placing a shadow over a candidate’s picture or blurring a background—would those images now need disclaimers?
• What about more common social media tools, such as photo filters? If a candidate uses AI-enhanced software to reduce wrinkles or grey hairs, will they run afoul of the Commission’s proposed rule if they do not use a disclaimer?

If taken literally, the proposed definition is so overbroad as to be self-defeating. If taken any other way, the proposed definition collapses in on itself, effectively becoming a standardless “know it when I see it” test, which fails to provide ordinary people with notice of what is actually required.

Citizens United cautioned that “[t]he First Amendment does not permit laws that force speakers to retain a campaign finance attorney, conduct demographic marketing research, or seek declaratory rulings before discussing the most salient political issues of our day.”[17] Yet, effectively requiring the retention of counsel or seeking agency guidance is precisely what the proposed rule would require.

b. The Proposed Definition of “Issue Advertising” is Impermissibly Vague

The definition of “political advertising” suffers from a similar infirmity. While the first part of the definition, “advertising that is made by or on behalf of a legally qualified candidate for public office,” is reasonably clear, the second part, defining “issue advocacy,” is not.

The proposed definition of “issue advertising” includes two types of messages: those “relating to any political matter” or “any controversial issue of public importance.” Both parts of the definition are problematic.

First, what is a “political matter?”

The term itself is undefined in the regulation. This sets it apart from the reference to “any political matter of national importance” in the statutory and regulatory provisions addressing licensees’ maintenance of a political file. Section 315 provides three examples of what constitutes a “political matter of national importance:” messages relating to “a legally qualified candidate,” “any election to Federal office,” or “a national legislative issue of public importance.”[18] Consistent with the ejusdem generis and noscitur a sociis canons of statutory construction, this specific enumeration of examples helps cabin in the potentially broad term “political matter of national importance.”[19]

The lack of accompanying examples renders the term “political matter” impermissibly vague. To see why, compare the Federal Election Commission and the Internal Revenue Service. The FEC generally has jurisdiction over public communications that expressly advocate the election of candidates for federal office or qualify as electioneering communications.[20] In contrast, the IRS relies on a much more expansive “facts and circumstances” approach that is not limited to only federal candidates in assessing “political campaign activity” by tax-exempt organizations.[21]

Does the Commission’s proposed definition of “political matter” track its political file rule? Does it track the FEC’s jurisdiction? The IRS’s definition of political campaign activity? Something else entirely? The Proposed Rule does not specify, leaving speakers to guess at their own peril.

Next, what does it mean to be “related to” a political matter?

Courts have noted that the phrase “related to” is “inevitably subject to criticism as overbroad since life, like law, is a seamless web, and all documents relate to all others in some remote fashion.”[22] This overbreadth is particularly problematic in the context of core First Amendment speech. For example, in Buckley v. Valeo, the Court only upheld language regulating speech “relative to” a clearly identified candidate after adopting a limiting construction to avoid constitutional infirmities.[23] There is no similar limiting construction proposed here. Thus, speakers are left to guess at what speech is covered by this provision.

Third, what is a “controversial issue of national importance?”

Again, unlike the Commission’s political file rule, the Proposed Rule provides no guidance or examples of what constitutes an “issue of national importance.” The political file rule is limited to candidate advocacy and national legislative issues. Does a similar limitation apply here? What about cross-cutting issues, such as advocating for or against state-level referenda relating to hot-button issues like abortion or immigration? The underlying issues themselves are plainly issues of national importance; is the discussion of state-level policy responses?

To make matters worse, the Commission introduces an inherently subjective term, “controversial.” What qualifies as a “controversial” issue? While some issues may be clear, others are not. What appears to be common sense to one person may strike against the deeply held belief of another. For example, not long ago, saying “there are only two genders” would not have raised many eyebrows. Today, at least one federal Court of Appeals has held school officials could reasonably conclude that a t-shirt with this message would be materially disruptive of a school environment.[24] How are speakers supposed to assess whether a topic is “controversial”? The Proposed Rule does not and cannot provide clear guidance.

The result is that people of ordinary intelligence must necessarily guess whether their proposed communications will trigger the disclosure requirements of the Proposed Rule. This is likely to chill both political speech and pure issue advocacy, in violation of the First Amendment.

III. The Regulatory Flexibility Analysis Ignores the Impact of the Proposed Rule on Campaigns and Advocacy Groups

The Regulatory Flexibility Analysis completely ignores candidates, campaigns, and advocacy groups. This is an egregious oversight that highlights the degree to which the Commission is institutionally ill-suited to regulate political speech.

While Commission licensees may be the group that is directly regulated by the Proposed Rule, they are not the ultimate speakers. Particularly when airing paid advertisements, they are generally conduits for the messages of others. It is the people who seek to air “political advertising” that are trying to exercise their First Amendment rights and engage in the marketplace of ideas. They are the ones who will ultimately be impacted by the Proposed Rule.

Campaigns and advocacy organizations are often small enterprises, and many are run by volunteers. It is hard to image a campaign or advocacy organization advocating on a “controversial” issue that could properly be considered “dominant” in its field. Thus, they are properly considered “small entities” for the purposes of a regulatory flexibility analysis.[25]

The Proposed Rule is unlikely to be costless for campaigns and advocacy groups. There are nearly 520,000 elected officials in the United States, and many of the elections for these offices feature more than one candidate.[26] Many thousands will almost certainly incur compliance costs if the Proposed Rule is to have any effect. For example, they will need to assess on the front end whether their communications meet the requirements for disclosure as “AI-generated content.” On the backend, particularly in a competitive political marketplace, they will incur costs addressing challenges from political opponents who seek to weaponize the Proposed Rules to make take-down requests or level other charges.

Particularly given the vagaries of the Proposed Rule, these costs are likely to be substantial in the aggregate. Yet, the Commission completely ignores them.

The Commission’s focus on cable companies, cable operators, and broadcasting companies makes sense in the context of the Commission’s work. After all, these are the parties the Commission typically regulates. But that is not primarily who the Commission would actually affect with its Proposed Regulation.

The Commission’s myopic focus on licensed entities merely emphasizes that it is not the appropriate authority to regulate the political space. This is not what the FCC is designed to do, and it is not what it is institutionally suited to do.

IV. The Proposed Rule Burdens Political Speech

Even taking the proposed rule on its own terms, there are a number of issues that are unexamined by the FCC. To wit, the FCC does not analyze the cumulative impact of combining the proposed spoken disclaimer with other required disclaimers on political advertisements. Radio advertisements that fall within the Federal Election Commission’s jurisdiction are required to include so-called “stand by your ad” provisions, which are spoken statements saying, for example, “I am candidate X, a candidate for [federal office], and I approve this message,” or for a “Y” organization, “Paid for by Y, [website address of Y]. Not authorized by any candidate or candidate’s committee. Y is responsible for the content of this advertising.”[27] The Proposed Rule would add an additional statement, “This message contains information generated in whole or in part by artificial intelligence.” Taken in insolation, each may appear trivial. But in a 30-second advertisement, these disclaimers quickly add up to a significant portion of a speaker’s time. This has a real cost to political speakers because it reduces their ability to convey their own message.

V. Conclusion

The Proposed Regulation strays far from the FCC’s core regulatory mission and intrudes on the prerogatives of Congress to adopt new laws. Worse, it disregards the potential impact on core First Amendment actors and utilizes vague and overbroad terms that will chill core First Amendment speech.

For the foregoing reasons, the Commission should reject the Proposed Rule and close this misguided effort.

Thank you very much for your time and consideration of our views.

Respectfully submitted,

Bradley A. Smith                                            Gary Lawkowski

Chairman                                                          Senior Fellow

[1] The Institute for Free Speech is a nonpartisan, nonprofit 501(c)(3) organization that promotes and defends the First Amendment rights to freely speak, assemble, publish, and petition the government.

[2] See 47 U.S.C. § 315(a); 47 C.F.R. § 73.1941(a).

[3] Biden v. Nebraska, 143 S. Ct. 2355, 2372 (2023).

[4] West Virginia v. EPA, 597 U.S. 697, 723 (2022) (quoting E. Gellhorn & P. Verkuil, Controlling Chevron Based Delegations, 20 Cardozo L. Rev. 989, 1011 (1999)).

[5] See Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 107 (2012).

[6] 52 U.S.C. § 30120(a); 11 C.F.R. § 110.11.

[7] 52 U.S.C. § 30120(d).

[8] 47 U.S.C. § 315(e)(2).

[9] 52 U.S.C. § 30106(b) (The FEC “shall have exclusive jurisdiction with respect to the civil enforcement of” the Federal Election Campaign Act of 1971, as amended, including its disclaimer provisions).

[10] Galliano v. USPS, 836 F.2d 1362, 1370 (D.C. Cir. 1988); see also Letter from Commissioner Sean J. Cooksey to the Honorable Jessica Rosenworcel (June 3, 2024), https://www.fec.gov/resources/cms-content/documents/FEC_Chairman_Cooksey_Letter_to_FCC_Chairwoman_Rosenworcel_June_3_2024.pdf (explaining that this proposal “would fall within the exclusive jurisdiction of the Federal Election Commission . . . ”).

[11] Supra n.5.

[12] 47 U.S.C.§ 303(r).

[13] 47 U.S.C. § 315(a); see also 47 C.F.R. § 73.1941(a).

[14] Id. at 324 (quoting Connally v. General Constr. Co., 269 U.S. 385, 391 (1926)).

[15] See, e.g., JibJab, This Land!, YouTube (Nov. 16, 2007), https://www.youtube.com/watch?v=z8Q-sRdV7SY; JibJab, Second Term, YouTube (Mar. 18, 2008), https://www.youtube.com/watch?v=hE8V22unwRo.

[16] See, e.g., Rob Christensen, New Adobe Lightroom AI Innovations Empower Everyone To Edit Like a Pro, Adobe Blog (Apr. 18, 2023), https://blog.adobe.com/en/publish/2023/04/18/new-adobe-lightroom-ai-innovations-empower-everyone-edit-like-pro.

[17] Supra n.18 at 324.

[18] 47 U.S.C. § 315(e)(1)(B); 47 C.F.R. § 73.1943(a)(2).

[19] See generally Fischer v. United States, 144 S. Ct. 2176, 2183–84 (2024) (“[T]he canon of noscitur a sociis teaches that a word is ‘given more precise content by the neighboring words with which it is associated.’ United States v. Williams, 553 U.S. 285, 294, 128 S. Ct. 1830, 170 L.Ed.2d 650 (2008). That ‘avoid[s] ascribing to one word a meaning so broad that it is inconsistent with’ ‘the company it keeps.’ Gustafson v. Alloyd Co., 513 U.S. 561, 575, 115 S. Ct. 1061, 131 L.Ed.2d 1 (1995). And under the related canon of ejusdem generis, ‘a ‘general or collective term’ at the end of a list of specific items’ is typically ‘‘controlled and defined by reference to’ the specific classes … that precede it.’ Southwest Airlines Co. v. Saxon, 596 U.S. 450, 458, 142 S. Ct. 1783, 213 L.Ed.2d 27 (2022) (quoting first Ali v. Federal Bureau of Prisons, 552 U.S. 214, 225, 128 S. Ct. 831, 169 L.Ed.2d 680 (2008); then Circuit City Stores, Inc. v. Adams, 532 U.S. 105, 115, 121 S. Ct. 1302, 149 L.Ed.2d 234 (2001)); accord, Bissonnette v. LePage Bakeries Park St., LLC, 601 U.S. 246, 252, 144 S. Ct. 905, 218 L.Ed.2d 204 (2024). These approaches to statutory interpretation track the common sense intuition that Congress would not ordinarily introduce a general term that renders meaningless the specific text that accompanies it. . . . The idea is simply that a general phrase can be given a more focused meaning by the terms linked to it.”).

[20] See generally supra n.17.

[21] See generally IRS, Restriction of Political Campaign Intervention by Section 501(c)(3) Tax-exempt Organizations, https://www.irs.gov/charities-non-profits/charitable-organizations/restriction-of-political-campaign-intervention-by-section-501c3-tax-exempt-organizations; Rev. Rul. 2007-41, 2007-25 I.R.B. (June 18, 2007), https://www.irs.gov/pub/irs-tege/rr2007-41.pdf.

[22] Freedom Watch, Inc. v. Dep’t of State, 925 F.Supp.2d 55, 61 (D.D.C. 2013) (cleaned up); see also Massachusetts v. Dep’t of Health & Human Servs., 727 F.Supp. 36, 36 n.2 (D.Mass. 1989).

[23] See Buckley v. Valeo, 424 U.S. 1, 40-44 (1976) (“We agree that in order to preserve the provision against invalidation on vagueness grounds, s 608(e)(1) must be construed to apply only to expenditures for communications that in express terms advocate the election or defeat of a clearly identified candidate for federal office.).

[24] See L.M. v. Town of Middleborough, 103 F.4th 854 (1st Cir. 2024).

[25] See 5 U.S.C. § 603(b)(3).

[26] Jennifer L. Lawless, Becoming a Candidate: Political Ambition and the Decision to Run for Office 33 tbl.3.1 (2012).

[27] See 52 U.S.C. § 30120(d); 11 C.F.R. §§ 110.11(c)(3)-(4).