PHOENIX â Just in time for campaign season: Another new state law designed to help voters figure out which images theyâre seeing are real and which are not.
Gov. Katie Hobbs has signed legislation requiring anything that purports to show a candidate within 60 days of an election to have âclear and conspicuous disclosureââ of whether it is instead a âdeep fake.ââ But it remains unclear whether the measure, SB 1359, is constitutional.
The move comes a week after Hobbs penned her approval to HB 2394 designed to protect candidates.
That bill, written by Rep. Alexander Kolodin, R-Scottsdale, allows those who are running for office to go to court to get a judicial declaration that the person in the picture, video or audio is not them. But on that one, there is no penalty, and a court cannot order the material removed from the web or airwaves.
Both measures are responses to artificial intelligence and the use of computers to create a picture, video or recording of someone else without their knowledge or consent, and to put words into their mouth.
âSay I wanted to make a deep fake of Chairman (Quang) Nguyen,ââ Sen. Frank Carroll told the House Judiciary Committee, which Nguyen, a Prescott Valley Republican, chairs.
He said the technology would be able to find various images and voice samples of Nguyen âbecause weâre all tied to the internetââ through everything from cell phones to social media.
âSo I can produce something that looks like Chairman Nguyen, except I can cause it to do something uncharacteristic or untrue about him,ââ said Carroll, a Sun City West Republican. âI can make him the most evil person in the world if I so describe it. But itâs not really him thatâs doing it.ââ
His SB 1359 signed by Hobbs goes a step beyond what Kolodin got enacted.
It spells out that anyone who distributes a âsynthetic media messageââ that the person knows is a âdeceptive and fraudulent deep fakeââ of a candidate on the ballot in the next 90 days must also provide âclear and conspicuous disclosureââ that it was generated using artificial intelligence.
It further defines âdeceptive and fraudulentââ as something the producer knows is false with the intent to injure the candidateâs reputation and is âintentionally calculated to mislead a reasonable person into concluding that a real individual said or did something that they did not say or do in reality.ââ
Marilyn Rodriguez, lobbyist for the American Civil Liberties Union of Arizona, told lawmakers she understands the concerns of those pushing new state laws.
However, âwe do not think itâs going to pass constitutional muster in court,ââ Rodriguez said.
The key, she said, is that the U.S. Supreme Court has ruled that speech â even false speech â is protected by the First Amendment. The fact that this âspeechââ is being produced through technology, Rodriguez argued, is legally irrelevant.
âAs a general matter, the First Amendment means the government has no power to restrict expression because of its message, its ideas, its subject matter or its content,ââ she said.
âSpeech cannot be censored simply because it is misleading,ââ Rodriguez said. âThe fact that a deep fake or recording depicts something that is untrue does not rob it of its First Amendment protections.ââ
She said there are constitutionally permissible restrictions on such speech. But Rodriguez said it has to fall within the parameters of situations such as harassment, fraud, extortion or âtrue threats.ââ
As to election matters, she said the courts have allowed some government regulation.
But here, too, Rodriguez said, there has to be some âcompelling governmental interest,ââ such as precluding messages that mislead people about voting requirements and procedures and actions designed to disenfranchise voters. Even then, she said, such restrictions kick in only when there is intentional misrepresentation of voter information â and not about candidates or their positions.
Kolodin, however, said those arguments miss the point.
âFalse speech is absolutely First Amendment protected,ââ he said. But the issue, said Kolodin, is not whether a statement is false but the fact that someone is being impersonated.
âItâs like if you signed my name on a loan document,ââ he said. âThatâs what weâre trying to target and prevent.ââ
âThat is an issue,ââ Rodriguez conceded. But she said guardrails are needed.
For example, Rodriguez noted, there already is a state law against criminal impersonation. But she noted it requires a showing it was done with the intent to defraud.
In fact, Rodriguez said the ACLU is supportive of SB 1078, another measure still making its way through the legislative process. That proposal by Sen. John Kavanagh, R-Fountain Hills, builds on that existing law and adds that it also would apply to using a computer-generated voice recording, image or video of another person âwith intent to defraud other persons or with intent to harass other persons.ââ
Rep. Justin Heap, R-Mesa, acknowledged there is a 2002 U.S. Supreme Court case that protects false speech. But he questioned whether the current justices would reach the same conclusion.
âWe have a completely new technology,ââ Heap said. âThis is video of someone with their voice speaking, saying something they didnât say. To me, it seems this brings in an entirely different dynamic.ââ
And, if nothing else, Heap said, a law like this would provide a basis for the Supreme Court to revisit its precedent.
âI actually hope this gets challenged because I think we need the court to give us some guidance on this and on new things,ââ he said.
Donât expect to start seeing the disclosure requirements right away. The law Hobbs signed will not take effect until at least the beginning of September, long after the stateâs July 30 primary election.
The other question is how effective the law will be in ensuring disclosure.
Carrollâs measure originally proposed a criminal penalty. But that was stripped out, replaced by a fine of $10 a day for the first 15 days and $25 daily after that.
