By Paul Coble, chair of Rose Law Group’s intellectual property & artificial intelligence department
On May 21, 2024, Gov. Katie Hobbs signed emergency House Bill 2394 into law in an attempt to curb the growing threat of digital impersonations created by generative-AI. The statute, now A.R.S. § 16-1023 (the “Anti-Deepfake Statute”), is aimed at protecting AZ citizens from personally damaging impersonations and prohibiting any impersonations of candidates appearing on AZ ballots in the near future. These are worthy, and indeed necessary, protections with the growing ubiquity of generative-AI tools.
There is just one problem: it will not work, at least not as intended. There are several gigantic gaps in the Anti-Deepfake Statute that make it ineffective for a large majority of the cases it was intended to help.
- Most Victims of Digital Impersonation Will Have No Actual Relief
The Anti-Deepfake Statute provides three categories of relief: (1) declaratory relief (2) injunctive relief; and (3) damages. For most causes of action–including impersonations of political candidates and those showing the subjects nude or engaged in crimes, sexual acts, and reputation-damaging activities–the only relief available is declaratory relief. (A.R.S. § 16-1023(F).) The last two categories–injunctive relief and damages–are available only for deepfakes of people who are not public figures, that are excessively sexual, and for which the publisher did not take “reasonable corrective action” after learning the impersonation was unauthorized. (A.R.S. § 16-1023(I).) The problem is that “declaratory relief,” by definition, does not stop the defendant from publishing or prevent them from further spreading the deepfake.
Declaratory relief arises when a potential defendant asks a court to declare their “rights, status, and other legal relations” before the potential plaintiff has brought the suit. (A.R.S. § 12-1831; A.R.S. § 41-1034.) They are often used, for example, to declare a patent invalid, determine the rights of parties to a contract, or clarify the application of a regulation. In the case of the Anti-Deepfake Statute, declaratory relief means only that a court would determine whether the media qualifies as a digital impersonation under the statute. In most cases that determination would carry some weight because declaratory judgments are considered final judgments (A.R.S. § 12-1831). Parties to final judgments are precluded from relitigating issues that have already been adjudicated against them, a legal concept known as estoppel. The Anti-Deepfake Statute, however, eliminates that benefit. Even if a court were to grant preliminary declaratory relief under the Anti-Deepfake Statute, none of the factual or legal determinations the court made could be considered later in that case or in any other proceeding. (A.R.S. § 16-1023(J).)
Under the Anti-Deepfake Statute, therefore, digital impersonations of public figures or those that do not include excessively sexual acts will only be entitled to a statement by the court saying as much–no injunction, no damages, no estoppel, no real relief.
- The Safe-Harbors for Publishers are Far Too Broad
Even those digital impersonations of non-public figures that are so offensive they could result in an injunction or damages are unlikely to see those results due to overly broad safe-harbor provisions. Injunctive or monetary relief may only be awarded if the publisher had “actual knowledge” at the time of publication that the media was a digital impersonation or fails to take “reasonable corrective action” within 21 days of having such actual knowledge. (A.R.S. § 16-1023(I)(3).) In other words, a publisher could bury its head in the sand about a heinous sexual image, learn that it is an unauthorized impersonation within minutes of its publication, and still wait nearly 3 weeks to do anything about it before incurring liability under the Anti-Deepfake Statute. So long as the publisher takes reasonable corrective action (including removing the impersonation and publishing a notice) within 21 days of having actual knowledge that it is an unauthorized impersonation, a deepfake victim is not entitled to any injunction or damages to prevent further harm.
In the age of digital propagation, 21 days is an inexcusable lifetime to allow the most violative category of impersonation to remain available. The authors of the Anti-Deepfake Statute understood that time is of the essence in deepfake cases–giving presiding judges only 2 days to rule on a motion for preliminary declaratory relief (A.R.S. § 16-1023(E))–yet allow a publisher more than 10 times that time to continue publishing a known impersonation without consequence. (A.R.S. § 16-1023(A)(1)(a).) Furthermore, the Anti-Deepfake statute does not apply to any media that the publisher conveys to its audience is a digital impersonation or that the authenticity of the media is disputed. A publisher may therefore continue distributing a deepfake impersonation, regardless of how damaging, in perpetuity, without any liability under the Anti-Deepfake Statute so long as the label the authenticity as disputed.
- Deepfakes Used In Paid Advertisements Can Easily Circumvent the Law
One of the main targets of the Anti-Deepfake Statute is digital impersonations of political candidates and preventing such deepfakes from sowing misinformation in elections. Most of those political deepfakes will be shared for free on social media, but some may end up in paid advertisements. If a digital impersonation is used in a paid advertisement, a cause of action under the AZ Anti-Deepfake Statute may only be brought against the one who “originated, ordered, placed or paid for the advertisement.” (A.R.S. § 16-1023(B).) Notably, this limitation excludes the publishers who were paid by 3rd-parties to distribute the deepfake advertisement. Actors with malintent could simply prepay a publisher to run an ad with a known deepfake and the publisher could not be enjoined or incur any damages under the AZ Anti-Deepfake Statute.
The Anti-Deepfake Statute takes aim at serious harms, but the statute as drafted does not adequately address those harms. It is critical to remedy these issues promptly to protect Arizona citizens from the increasing threat of digital impersonation.
