在本页4 分钟阅读
概述
More than 20 US states have adopted some version. These laws matter because a convincing fake robocall or video released days before a vote can mislead voters before anyone debunks it. They also run into the strong First Amendment protection that US law gives political speech.
深入探讨
Texas passed the first deepfake election law in 2019, making it a crime to create and publish a deceptive deepfake video within 30 days of an election with intent to injure a candidate. California followed the same year with a law that includes a safe harbor for content that carries a disclosure. After generative AI tools spread in 2023, many more states acted, including Minnesota, Michigan and Washington. The laws follow two broad models. Prohibition laws ban distributing materially deceptive synthetic media of a candidate within a set window before an election. Disclosure laws allow the content if it carries a clear label. Windows vary by state, commonly from about 30 to 120 days, and most laws exempt satire, parody and news reporting. Federal action has been narrower. After the Biden robocall, the FCC ruled in February 2024 that AI-generated voices count as "artificial" under the Telephone Consumer Protection Act, which made such robocalls illegal without prior consent. The FCC fined consultant Steve Kramer $6 million; a New Hampshire jury later acquitted him of related state criminal charges in 2025. The FCC also proposed AI disclosure rules for broadcast political ads, but did not finalize them. The Federal Election Commission declined to write new rules and said instead that its existing ban on fraudulent misrepresentation applies whatever technology is used. Courts have been skeptical of broad bans. In Kohls v. Bonta a federal judge blocked California's 2024 law against deceptive election content, finding it swept in protected speech such as parody. A companion law requiring large platforms to remove such content was later also struck down. Minnesota's law was challenged as well. A common misconception is that false political speech is unprotected. In United States v. Alvarez (2012), the Supreme Court declined to treat lies as a category of speech outside First Amendment protection.
战略影响
风险与安全
灾难性和日常的人工智能危害都取决于谁了解风险以及谁能够采取行动。
更清晰的判决
公众和专业素养决定强有力的安全政策在政治上是否可行。
打破炒作
清晰的解释可以减少炒作、实验室公关和模糊道德剧场的影响。
The Future of Deepfake Election Laws and Political Ad AI Disclosure
More states are likely to adopt or amend these laws, shifting toward disclosure requirements, which courts have treated more favorably than outright bans. Litigation will keep testing where regulation of deception ends and censorship of parody begins. Federal legislation has been proposed repeatedly but has not passed. Platform policies and the FCC's robocall ruling remain the main national tools. Provenance standards such as C2PA may make it easier to show that content is authentic, but they cannot prove that unlabeled content is fake.
现实世界的实施
Before New Hampshire's January 2024 primary, a robocall used an AI clone of President Biden's voice to tell Democrats not to vote. The FCC later fined the political consultant behind it $6 million, and the telecom carrier that transmitted the calls agreed to a $1 million settlement.
A campaign in a state with a disclosure law runs an ad showing an AI-generated scene of its opponent. It adds a clear label stating that the content was generated or manipulated with AI, which brings it within the law's safe harbor.
A parody video that altered Kamala Harris's voice spread widely in 2024. A new California law aimed at such content was challenged in court and blocked by a federal judge that October.
An advertiser buying election ads on Google or Meta must disclose when an ad contains realistic synthetic content, and the platform then shows a label to viewers.
风险与防护栏
将存在风险视为科幻小说，同时能力复合。
混淆了表面产品安全与高度自治下的对准。
只给非英语和非专业观众留下低质量的资源。
实施路线图
单独的产品危害、误用和失控/失调风险。
询问哪些证据会改变您对时间表和严重性的看法。
比起营销主张，更喜欢主要来源和具体评估。
确定一条行动路径：职业、政策、资金或技能——而不仅仅是意识。
不断探索
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Deepfake Election Laws and Political Ad AI Disclosure quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
常见问题
What is Deepfake Election Laws and Political Ad AI Disclosure?
Deepfake election laws are state statutes that ban, or require labels on, deceptive AI-generated audio, images or video of candidates in the weeks before an election. More than 20 US states have adopted some version. These laws matter because a convincing fake robocall or video released days before a vote can mislead voters before anyone debunks it. They also run into the strong First Amendment protection that US law gives political speech.
Which state passed the first deepfake election law in 2019?
Texas's 2019 law made it a crime to create and publish a deceptive deepfake video within 30 days of an election with intent to injure a candidate. California passed its own law later that year.
What did the FCC rule in February 2024 after the Biden robocall?
Classing AI voices as "artificial" under the TCPA made AI-voice robocalls illegal without prior consent.
What is the main difference between prohibition laws and disclosure laws?
Prohibition laws ban deceptive synthetic media within a window before an election. Disclosure laws allow it with a clear AI label.
What happened in Kohls v. Bonta?
The court found the law burdened protected expression such as parody and blocked it. A companion platform law was later struck down as well.
How did the Federal Election Commission respond to calls for AI rules?
The FEC chose to interpret existing law as covering AI-enabled fraud rather than create a new AI-specific rule.
继续学习
相关指南
为此主题精选的更多指南