Attorney General Opinions and Advisory Letters

OPINION OF RAÚL TORREZ

July 3, 2025

Opinion No. 2025-09

Attorney General

To: Re:

1.

2.

3.

4.

1.

2.

Honorable Michelle Lujan Grisham, Governor of the State of New Mexico Attorney General Opinion – Facial Constitutionality of HB 182

Questions Is the Campaign Reporting Act’s definition of “artificial intelligence” (AI) unconstitutionally vague?

Does the Act’s disclaimer requirement violate the First Amendment’s standard of exacting scrutiny given that it does not apply to media produced with the depicted individual’s consent?

Is the Act’s disclaimer requirement unconstitutionally overbroad in that it applies to advertisements that “reasonably constitute[] satire or parody?”

Does the Act’s amendment permit private individuals and organizations to prosecute criminal violations of the Act?

Summary of Answers No. Facial vagueness challenges to statutes at the pre-enforcement stage are limited. The current definition of AI can reasonably be understood on its face by a person of ordinary intelligence as applying to generative AI applications capable of producing realistic deepfakes, and the law does not appear to be facially overbroad.

No. The limited scope of the law and its exception for consensual depictions are not subject to an underinclusiveness test under the First Amendment.

3.

4.

Yes. Satire and parody, by definition, are not likely to mislead a reasonable viewer or listener into believing that they reflect actual facts. There is thus an insufficient justification for requiring a disclaimer for these types of advertisements.

No. The statute does not authorize criminal enforcement by private individuals or special interest groups.

Background On March 6, 2024, Governor Michelle Lujan Grisham signed into law an amendment to the Campaign Reporting Act, NMSA 1978, §§ 1-19-26 through -37 (1979, as amended through 2024). The amendment addresses the problem of AI-generated deepfakes and their potential to mislead voters in relation to candidates or ballot issues. The amendment has three primary components: (1) it requires a disclaimer for advertisements that contain “materially deceptive media” indicating that they were created using AI; (2) it provides that it is a violation of the Act to knowingly distribute materially deceptive media within ninety days of an election with the intent to change voting by misleading voters in a way reasonably likely to result in a change to voting unless accompanied by the disclaimer; and (3) it establishes criminal liability for willfully and knowingly violating the distribution proscription. NMSA 1978, §§ 1-19-26.4 (2024), -26.8 (2024). Violations of the amended sections of the Act can result in civil penalties and injunctive relief, and a knowing and willful violation of the distribution provision is a misdemeanor for a first offense and a fourth degree felony for a second offense.

Before its enactment, the amendment was the subject of opposition questioning its constitutionality under the First Amendment. Following the amendment’s enactment, the Governor requested a formal opinion from the Attorney General addressing several potential constitutional infirmities. The request does not refer to any specific application of the Act and thus focuses on the language of the amendment on its face.

Analysis 1. Is the Act’s definition of “artificial intelligence” unconstitutionally vague? Statutes are generally entitled to a strong presumption of constitutionality. See State v. Laguna, 1999-NMCA-152, ¶ 24, 128 N.M. 345. A claim of vagueness is typically analyzed according to the particular facts of a case. See State v. Luckie, 1995-NMCA-075, ¶ 6, 120 N.M. 274. “For a host of good reasons, courts usually handle constitutional claims case by case, not en masse.” Moody v. NetChoice, LLC, 144 S. Ct. 2383, 2397 (2024). Although statutes may be challenged on their face, such claims are “disfavored” and “hard to win.” Id. at 2397, 2409.

By and large, facial vagueness challenges must establish that a law is impermissibly vague in all its applications. See N.M. Petroleum Marketers Ass'n v. N.M. Env't Improvement Bd., 2007-NMCA-060, ¶ 16, 141 N.M. 678. However, when a law implicates the First Amendment such that it could potentially result in a chilling effect on constitutionally protected speech, the Supreme Court applies a special overbreadth test to facial challenges: “[A] law may be invalidated as overbroad if ‘a substantial number of its applications are unconstitutional, judged in relation to the

statute’s plainly legitimate sweep.’” United States v. Stevens, 559 U.S. 460, 473 (2010) (quoting Washington State Grange v. Washington State Republican Party, 552 U.S. 442, 449 n.6 (2008)). To succeed in making such a claim, a party must usually “describe the instances of arguable overbreadth.” Washington State Grange, 552 U.S. at 449 n.6.

In the 2024 amendment, the Legislature defined “artificial intelligence” as “a machine-based or computer-based system that through hardware or software uses input data to emulate the structure and characteristics of input data in order to generate synthetic content, including images, video or audio.” NMSA 1978, § 1-19-26(C) (2024). Opponents of the law before its enactment suggested that this definition fails to identify the type of “system” to which it applies and does not specify whether it applies to such applications as Snapchat, Instagram, and Adobe Photoshop. These suggestions, however, do not come close to establishing impermissible vagueness in all the statute’s applications as would be required for a facial challenge.

The definition of AI does not appear in the substantive provisions of the amendment and instead informs the definition of “materially deceptive media,” which means an

image, video or audio that: (1) depicts an individual engaged in conduct or speech in which the depicted individual did not engage; (2) was published, disseminated, distributed or displayed to the public without the consent of the depicted individual; and (3) was produced in whole or in part by using artificial intelligence.

Section 1-19-26(S). When the Legislature defines a term, the express definition in the statute supplies “the Legislature’s intended meaning.” Morris v. Brandenburg, 2016-NMSC-027, ¶ 15, 376 P.3d 836. In addition, “[a] word or phrase that has acquired a technical or particular meaning in a particular context has that meaning if it is used in that context.” NMSA 1978, § 12-2A-2 (1997). “[P]erfect clarity and precise guidance have never been required even of regulations that restrict expressive activity.” Ward v. Rock Against Racism, 491 U.S. 781, 794 (1989).

New Mexico is not alone in enacting laws designed to curb the harmful impact of deepfakes on elections. These laws target advanced and complex generative AI that “mak[es] it difficult for the average person to detect the falsity of a deepfake.” Kohls v. Ellison, 2025 WL 66765, at *1 (D. Minn. Jan. 10, 2025). In the context of the broader definition of “materially deceptive media,” and in light of the technical meaning of AI in the deepfake context, it appears that the statute provides sufficient notice of its scope to an average person of ordinary intelligence and provides minimum guidelines for enforcement so as not to be unconstitutionally vague on its face. See State v. Jacquez, 2009-NMCA-124, ¶ 6, 147 N.M. 313 (describing two ways in which a statute might be void for vagueness).

Because the 2024 amendment implicates the First Amendment, the Supreme Court’s special overbreadth analysis applies in the context of a facial challenge. But the suggestion that the definition of AI could conceivably apply to images created with common social media applications fails to describe specific instances of arguable overbreadth. Moreover, this hypothetical challenge to the statute does not establish a substantial number of unconstitutional applications of the statute in relation to its overall legitimate scope. The below analysis of the law in the abstract does not

reveal an obvious constitutional infirmity as the issue has been presented in the request for an opinion.

The 2024 amendment to the Act applies to advertisements “referring to a candidate or ballot question,” Section 1-19-26(A), that is, it applies to political speech. As a general matter, “political speech must prevail against laws that would suppress it,” and restrictions on political speech are subject to strict scrutiny under which there must be a compelling governmental interest and the use of the least restrictive means to effectuate that interest. Citizens United v. FEC, 558 U.S. 310, 340 (2010). “The freedom to speak one’s mind is not only an aspect of individual liberty – and thus a good unto itself – but also is essential to the common quest for truth and the vitality of society as a whole.” Bose Corp. v. Consumers Union of United States, Inc., 446 U.S. 485, 503-04 (1984). “The sort of robust political debate encouraged by the First Amendment is bound to produce speech that is critical of those who hold public office . . . .” Hustler Magazine, Inc. v. Falwell, 485 U.S. 46, 51 (1988).

In some states, like California, deepfake statutes prohibit the distribution of materially deceptive media. At least one court has issued a preliminary injunction against such a law based on a finding that the plaintiff was likely to succeed in establishing a failure to narrowly tailor the law to the government’s interest in protecting elections. Kohls v. Bonta, 752 F. Supp. 3d 1187, 1193-96 (E.D. Cal. 2024).

New Mexico’s law, unlike the one in California, does not prevent the creation, production, purchase, or distribution of materially deceptive media. Instead, it requires that such advertisements contain a disclaimer. Section 1-19-26.4(D), -26.8(B). Statutes that create disclaimer requirements for political speech are subject to a lower standard of exacting scrutiny, rather than strict scrutiny, because such laws “impose no ceiling on campaign-related activities and do not prevent anyone from speaking.” Citizens United, 558 U.S. at 366-67 (quotation marks and quoted authority omitted). Exacting scrutiny requires that the disclaimer requirement bear a substantial relationship to a sufficiently important governmental interest, and it requires that the disclaimer requirement be “narrowly tailored to the government’s asserted interest.” Americans for Prosperity Found. v. Bonta, 594 U.S. 595, 608 (2021).

Section 1-19-26.4(D) requires a disclaimer that an image, video, or audio advertisement “has been manipulated or generated by artificial intelligence.” The disclaimer need only be readable or audible. Section 1-19-26.4(E). The disclaimer prevents viewers or listeners from being misled into believing that an advertisement depicts conduct or speech genuinely engaged in by the depicted individual. In this way, the disclaimer serves the “sufficiently important” governmental interest in protecting the voting public from misinformation or disinformation about an election. The disclaimer requirement does not appear to create an undue burden on an advertiser. See Buckley v. Valeo, 424 U.S. 1, 68 (1976) (evaluating “the extent of the burden” that disclosure requirements “place on individual rights”). The disclaimer is brief, non-obtrusive, factual, and provides information that is not otherwise available to a viewer or listener. Indeed, the inability of an average viewer or listener to be able to detect a false AI-generated depiction makes it less susceptible to being remedied with the truth in the marketplace of ideas. See United States v. Alvarez, 567 U.S. 709, 727 (2012) (plurality opinion) (“The remedy for speech that is false is speech that is true.”). The statute thus appears to be narrowly tailored to the government interests it serves and not overly

burdensome. In short, the law does not appear to have the type of impermissible overbreadth that would support a facial challenge to its constitutionality.

2.

Does the Act’s disclaimer requirement survive exacting scrutiny given that it does not apply to media produced with the depicted individual’s consent?

The Legislature, as part of the definition of materially deceptive media, required that the media be “published, disseminated, distributed or displayed to the public without the consent of the depicted individual.” Section 1-19-26(S). Opponents of the 2024 amendment to the Act argued that it would not survive First Amendment scrutiny because it excludes consensual depictions, thereby preventing the law from accomplishing its aim of protecting voters from deceptive advertising. The Act shows, however, that the Legislature had a narrower aim and did not design these provisions to target all deceptive advertising. The Legislature intended to prevent a person’s likeness from being used for a portrayal that the person had not approved. In other words, the Legislature viewed a lack of consent as an integral component of the deception and falsity of the media. Consensual portrayals may be artificially generated, but they are less likely to mislead voters about the depicted individual’s thoughts and views as expressed through the adopted words and actions of the individual’s likeness. Indeed, the content of the disclosure requirement reflects this more limited purpose. The Legislature did not require a disclaimer that a particular advertisement is potentially deceptive or misleading; instead, the disclaimer must only indicate that the advertisement “has been manipulated or generated by artificial intelligence.” The disclaimer therefore merely alerts a viewer or listener that the depiction of the individual is not genuine, a disclaimer that would have less impact for consensual renderings. It thus appears that the disclaimer is substantially related to an important governmental interest and is narrowly tailored to the interest.

Moreover, the concern about excluding consensual depictions from the law raises an issue of underinclusiveness or underbreadth, essentially arguing that the Legislature should have targeted all artificial and deceptive advertising. The United States Supreme Court has made it clear, however, that the First Amendment does not impose “an ‘underinclusiveness’ limitation.” R.A.V. v. St. Paul, 505 U.S. 377, 387 (1992). The Legislature is not required to solve all problems, or even the whole of any single problem, at one time. See Denver Area Telecomm. Consortium, Inc. v. F.C.C., 518 U.S. 727, 757 (1996); see also id. at 836-37 (Thomas, J., concurring in the judgment in part and dissenting in part). The Legislature’s decision to exclude consensual depictions from the reach of the Act does not violate the First Amendment.

3.

Is the Act’s disclaimer requirement unconstitutionally overbroad in that it applies to advertisements that “reasonably constitute[] satire or parody”?

Parody and satire are protected by the First Amendment. See Cliffs Notes, Inc. v. Bantam Doubleday Dell Publ’g Grp., Inc., 886 F.2d 490, 493 (2d Cir. 1989). “Satire’s unifying element is the use of wit to expose something foolish or vicious to criticism. A ‘parody’ is to the same effect: the style of an individual or work is closely imitated for comic effect or in ridicule.” Farah v. Esquire Mag., 736 F.3d 528, 536 (D.C. Cir. 2013) (quotation marks and quoted authority omitted). Satire and parody do not give rise to liability for deliberate falsehoods because they cannot be reasonably understood or interpreted as stating actual facts. See id. at 535-37.

The Legislature expressly provided that “an advertisement that reasonably constitutes satire or parody” does not violate Section 1-19-26.4 “if the advertisement includes a disclaimer consistent with the requirements provided in Subsection D of this section.” Section 1-19-26.4(G)(3). It is worth noting that the Legislature’s intent with respect to parody and satire is unclear; the statute specially designates satire and parody as non-violative of the statute but, in tension with this designation, requires the disclaimer that is required of all other advertisements with materially deceptive media. Read literally, Subsection G(3) merely repeats Subsection D for a specific type of advertisement. Although a “statute must be construed so that no part of the statute is rendered surplusage or superfluous,” State v. Javier M., 2001-NMSC-030, ¶ 32, 131 N.M. 1 (quotation marks and quoted authority omitted), it will be presumed that the Legislature required satire and parody to comply with the disclaimer requirement applicable to other advertisements for purposes of responding to the present inquiry.

As with Section 1-19-26.4’s application to other advertisements, the statute does not prohibit the creation, production, purchase, or distribution of parody and satire generated with AI; instead, the statute requires the use of a disclaimer. This provision is therefore reviewed under the exacting scrutiny standard discussed above. But the statute’s treatment of satire and parody differs from other advertisements in terms of the nexus between the government’s interests and the disclaimer requirement. Section 1-19-26.4 and Section 1-19-26.8 are designed to protect against misleading advertisements. See Section 1-19-26.8(A) (requiring not only an intent to mislead voters but also that the distribution of the advertisement be “reasonably likely to cause that result”). By its nature, however, “an advertisement that reasonably constitutes satire or parody” is not likely to mislead a reasonable viewer or listener. The disclaimer is thus not substantially related to the governmental interests it is designed to protect and is not narrowly tailored for this purpose. Cf. Kohls v. Bonta, 752 F. Supp. 3d at 1196-97 (concluding that a disclaimer requirement for satire and parody was likely unconstitutional because it was unduly burdensome and not narrowly tailored). “Despite their sometimes caustic nature, from the early cartoon portraying George Washington as an ass down to the present day, graphic depictions and satirical cartoons have played a prominent role in public and political debate.” Hustler Mag., 485 U.S. at 54. Given the importance of satire and parody to the marketplace of ideas, this part of the statute is likely unconstitutional on its face.

4.

What sort of enforcement actions does the Act authorize private individuals and organizations to bring?

Before the 2024 amendment, a violation of the Act could result in civil penalties, injunctive relief, and criminal penalties. The Legislature provided the State Ethics Commission with the authority to “institute a civil action in district court” for a violation of the Act seeking injunctive relief or civil penalties, or both. NMSA 1978, § 1-19-34.6 (2021). A knowing and willful violation of the Act is a misdemeanor that “may be enforced by the attorney general or the district attorney where the candidate resides, where a political committee has its principal place of business or where the violation occurred.” NMSA 1978, § 1-19-36(B) (2021). The Secretary of State “may refer a matter to the state ethics commission for a civil injunctive or other appropriate order or to the attorney general or a district attorney for criminal enforcement.” NMSA 1978, § 1-19-34.4 (2021). The Act did not mention private enforcement.

The Legislature amended the Act in 2024 by specifying that the distribution of materially deceptive media is a violation of the Act and that a willful and knowing violation is a crime. Section 1-19-26.8(A), (B). The Legislature further provided for enforcement of this section of the Act

by any of the following: (1) the attorney general; (2) a district attorney; (3) a depicted individual who is falsely represented; (4) a candidate for office who has been injured or is likely to be injured by the distribution of materially deceptive media; or (5) any organization that represents the interests of voters who are likely to be misled by the distribution of materially deceptive media.

Section 1-19-26.8(E). The Legislature, however, did not amend or modify the sections of the Act that give the State Ethics Commission the authority to seek civil penalties and the Attorney General and the District Attorney the authority to enforce criminal violations of the Act.

The Attorney General is the law officer of the State and has the authority to prosecute or defend civil or criminal actions in which the State is a party or may be interested. NMSA 1978, § 8-5-2(B) (1975). The District Attorneys are the law officers of the State and the counties within their districts, N.M. Const. art. VI, § 24, and they have the authority to prosecute or defend civil and criminal actions in which the State or a county of their district is a party or may be interested. NMSA 1978, § 36-1-18(A)(1) (2001). Consistent with these provisions, the Legislature has directed that

no one shall represent the state or any county thereof in any matter in which the state or county is interested except the attorney general, his legally appointed and qualified assistants or the district attorney or his legally appointed and qualified assistants and such associate counsel as may appear on order of the court, with the consent of the attorney general or district attorney.

NMSA 1978, § 36-1-19(A) (1985). Courts lack jurisdiction over any attempted criminal prosecution by a private individual initiated without the consent of the Attorney General or the District Attorney and without the court’s approval. State v. Baca, 1984-NMCA-096, ¶¶ 8-10, 101 N.M. 716.

If Section 1-19-26.8(E) were construed to authorize private prosecutions, it would conflict with Section 36-1-19 and the constitutional and statutory authority of the Attorney General and the District Attorneys to represent the State in criminal proceedings. Such a construction of the statute would be contrary to established rules of statutory construction. In the interpretation of a statute, the Legislature is presumed to be “aware of existing law when it undertakes to amend its own statutes.” State v. Clah, 1997-NMCA-091, ¶ 11, 124 N.M. 6. Repeals by implication are disfavored and require clear and explicit language replacing an existing statute. Ferlic v. Mesilla Valley Reg’l Dispatch Auth., 2025-NMSC-___¶ 19, ___ P.3d ___ (S-1-SC-40162, April 21, 2025). Further, “[i]f statutes appear to conflict, they must be construed, if possible, to give effect to each.” NMSA 1978, § 12-2A-10(A) (1997). Finally, statutes are construed to avoid constitutional questions whenever possible. Lovelace Med. Ctr. v. Mendez, 1991-NMSC-002, ¶ 12, 111 N.M. 336.

Guided by these principles, it is apparent that Section 1-19-26.8(E) does not grant private individuals the authority to appear for the State for purposes of seeking civil or criminal penalties. The Legislature did not express any intent to modify Section 36-1-19, Section 1-19-34.6, or Section 1-19-36. Moreover, Section 1-19-26.8(E) does not refer expressly to civil or criminal penalties but expressly refers to injunctive relief. Interpreting the statute as a whole and in pari materia with other statutes, there is little question that the Legislature intended to allow the private individuals listed in Section 1-19-26.8(E) to seek injunctive relief for a violation of this section but not to seek civil or criminal penalties on behalf of the State.

* * * Please note that this opinion is a public document and is not protected by the attorney-client privilege. It will be published on our website and made available to the general public.

RAÚL TORREZ ATTORNEY GENERAL

/s/ James Grayson James Grayson Chief Deputy Attorney General