Deepfake Law News Today: The 2026 Legal Tracker
Tracking deepfake laws across the U.S. federal government, all 50 states, and major jurisdictions worldwide: what’s in force, what courts have struck down, and what’s coming next. Updated monthly.
Last updated: July 2026In 2026, deepfakes are no longer a legal gray zone. The United States has a dedicated federal law in full effect, more than forty states have legislated, and mandatory labeling and criminal rules from the EU, South Korea, China, Canada, and India are all landing this year. Meanwhile, the first wave of cases is testing the new laws: some defendants have pleaded guilty, one walked free at trial, one law exposed its first loophole three weeks after taking effect, and others have been struck down entirely.
This page tracks all of it in language anyone can read: the latest developments, the real status of federal and state laws, live test cases from the 2026 midterms, global legislation, and what to do if you become a victim yourself. Updated once a month.
Latest Updates
- August 2, 2026 (upcoming)
The EU AI Act’s Article 50 transparency obligations take effect: deploying deepfake content in the EU requires clear disclosure that it is AI-generated, with fines up to €15 million or 3% of global annual turnover. The machine-readable watermarking duty for AI systems already on the market before August 2 is deferred to December 2. The same day, California’s SB 942 (AI transparency law) also takes effect — deliberately synced with the EU.
- July 24, 2026
An Arkansas family sued xAI, alleging its Grok model was used by a photographer to generate child sexual abuse material of minors. The photographer already faces 200 state charges. It is one of the first family lawsuits in the U.S. aimed directly at an AI platform, and the family’s lawyers say more suits are weeks away.
- July 22, 2026
Tennessee’s political deepfake disclosure law (HB 1513) exposed its first loophole just three weeks after taking effect: in the 6th Congressional District’s Republican primary, two candidates ran undisclosed AI attack ads against each other — and both ads were legal, because the law covers only state-level races, not federal congressional campaigns.
- July 18, 2026
Canada’s Bill C-16 (Protecting Victims Act) came into force: creating and distributing non-consensual sexual deepfakes is now a crime, threatening to distribute carries up to 10 years, and platforms must remove reported content within 48 hours.
- June 18, 2026
The NO FAKES Act cleared the Senate Judiciary Committee unanimously — the first time the bill creating a federal “digital replica” right has made it out of committee. It now awaits a floor vote.
- June 11, 2026
The DOJ and DHS seized the domain of deepfake porn site CFAKE.com under the TAKE IT DOWN Act, coordinating with French police, who made an arrest in Nice. It was the law’s first use in a cross-border takedown.
- May 19, 2026
The TAKE IT DOWN Act’s platform obligations took effect: platforms across the U.S. must now remove reported non-consensual intimate images (including AI forgeries) within 48 hours. The same month, the FTC sent warning letters to 15 major platforms and 12 “nudify” websites, and federal prosecutors unsealed the first two criminal cases under the law.
- April 7, 2026
Ohio man James Strahler II pleaded guilty, becoming the first person in the country convicted under the TAKE IT DOWN Act: he used more than 100 AI tools to generate and publish forged intimate images of multiple women. Sentencing is pending.
- January 13, 2026
The DEFIANCE Act passed the Senate unanimously (for the second time), propelled by that month’s controversy over Grok being used to mass-generate “undressed” images of real people. The bill would give victims a federal civil claim with statutory damages starting at $150,000. It remains stalled in the House.
Federal Deepfake Laws
The federal picture: one law in force, two in the pipeline. The TAKE IT DOWN Act, already in effect, covers non-consensual intimate images. Of the two still in Congress, one would give victims the right to sue for damages (DEFIANCE), and one would give everyone property rights in their own face and voice (NO FAKES).
TAKE IT DOWN Act: In Force — and Biting
Signed May 19, 2025 — full name: Tools to Address Known Exploitation by Immobilizing Technological Deepfakes on Websites and Networks Act — after passing the House 409–2 and the Senate unanimously. It does two things:
First, criminalization. Knowingly publishing intimate images of an identifiable real person online — whether authentic or AI-generated (the law’s term is “digital forgery”) — carries up to 2 years in prison for adult victims and up to 3 years when minors are involved. This part took effect the day it was signed.
Second, platform duties. Since May 19, 2026, platforms must operate a reporting channel and remove flagged images — plus known identical copies — within 48 hours of a valid request. The Federal Trade Commission enforces, with civil penalties up to $53,088 per violation and no cap on the number of violations.
The law is no longer text on paper. In April 2026, Ohio’s James Strahler II pleaded guilty in the first TAKE IT DOWN conviction (sentencing pending); in May, the FTC sent warning letters to 15 platforms — including Amazon, Meta, Microsoft, TikTok, and X — and 12 “nudify” websites, and opened a victim complaint portal at TakeItDown.ftc.gov; in June, the DOJ used the law to seize CFAKE.com. Meanwhile, the Electronic Frontier Foundation and others continue to press First Amendment and abuse concerns.
DEFIANCE Act: Unanimous in the Senate, Stuck in the House
If TAKE IT DOWN is the state punishing offenders, DEFIANCE is victims suing for themselves. It creates a federal civil claim for victims of non-consensual sexual deepfakes: statutory damages starting at $150,000, rising to $250,000 where the conduct involves sexual assault, stalking, or harassment — plus actual damages, punitive damages, and attorney’s fees, with a 10-year window to sue.
The Senate passed it unanimously on January 13, 2026, driven directly by that month’s controversy over Grok generating sexualized images of real people. But that was the second time the Senate passed it unanimously: in 2024, it died in the House. The same script is playing out again — the bill has sat in the House Judiciary Committee since May 2025, with no action in over a year.
NO FAKES Act: Your Face and Voice Are Becoming Property
On May 20, 2026, the bill was introduced for the fourth time (counting from the 2023 discussion draft), and on June 18 it cleared the Senate Judiciary Committee unanimously — the furthest it has gone in three years.
What it would do is big: create a new federal intellectual property right giving every person control over “digital replicas” of their voice and likeness. Unauthorized creation and distribution would carry liability, with statutory damages up to $750,000 per work. The right survives death for 70 years; licensing while alive is capped — 10 years for adults, 5 for minors. The 2026 version adds a DMCA-style notice-and-takedown system with counter-notices, giving platforms a safe harbor if they follow it; news, documentaries, commentary, and parody are expressly exempt.
Its relationship with state law is carefully drawn: it preempts state digital-replica laws passed after January 2, 2025, but preserves earlier ones — so Tennessee’s 2024 ELVIS Act, the first law in the country to write “voice” into personality property rights, survives.
Support and resistance are equally clear: the RIAA and the AFL-CIO back it; Senators Lee, Cruz, and Schmitt voted yes while flagging First Amendment reservations; GovTrack currently gives it a 27% chance of passage. It started later than DEFIANCE but may travel further.
Prologue: The FCC and That “Biden” Robocall
The federal government’s first strike against deepfakes actually predates all of these bills. In February 2024, the FCC ruled that AI-cloned voices count as “artificial voices” under the robocall law (TCPA), making them illegal without prior consent. The trigger was the fake Biden robocall before the New Hampshire primary — organizer Steven Kramer was fined $6 million by the FCC, but in June 2025 a New Hampshire jury acquitted him of all criminal charges, felonies and misdemeanors alike; as of late 2025, the fine remained unpaid. A regulatory fine on one side, a criminal acquittal on the other — the case became the emblem of law struggling to catch up with technology. Full case study on our media literacy page.
The Undercurrent: The Federal–State Preemption Fight
Before reading the state section, keep one undercurrent in mind. In July 2025, the Senate voted 99–1 to strip a proposed ten-year freeze on state AI regulation — Congress refused to freeze state law. Five months later, the White House took the litigation route instead: a December executive order directed the DOJ to form an AI Litigation Task Force to sue “onerous” state AI laws (child-safety laws expressly exempt). In April 2026, the DOJ entered the fray for the first time, joining xAI’s challenge to Colorado’s AI Act — and days later, a court paused that law’s enforcement. So far, the task force hasn’t touched a single deepfake-specific law. But its existence means every state law below now faces one more variable beyond the courts.
State Deepfake Laws
The federal government only laid the floor; the fine-grained rules live at the state level. In 2025, all 50 states introduced AI-related bills — more than 1,080 in total, of which 301 dealt with deepfakes, and 68 sexual-deepfake bills became law (MultiState tally). Today the state landscape splits into three blocks: elections, sexual images, and child protection.
Election Deepfake Laws: 31 States
As of June 2026, 31 states have laws on deepfakes in elections (Public Citizen tracker). The vast majority take the “disclosure” route — AI content isn’t banned, but it must be labeled. The few that tried time-limited bans are losing in court, one after another.
| State | Law | Core requirement | Status |
|---|---|---|---|
| Texas | SB 751 (2019, first in the nation) | Bans deepfake videos intended to harm a candidate within 30 days of an election | In force |
| California | AB 2839 (2024) | Banned “materially deceptive” election content | Struck down in its entirety, Aug 2025 (First Amendment); state appealing |
| California | AB 2655 (2024) | Required large platforms to remove election deepfakes | Struck down the same day — on Section 230 preemption grounds |
| California | AB 2355 (2024) | AI disclosure in political ads | In force, unchallenged |
| Minnesota | §609.771 (2023) | Ban within 90 days of an election | In force; two lawsuits ongoing (appeals court kept it alive on procedural grounds, Feb 2026) |
| Michigan | HB 5141–5145 (2023) | AI disclosure within 90 days of an election; $250–$1,000 fines; satire exempt | In force |
| Oregon | SB 1571 (2024) | AI disclosure in campaign materials; up to $10,000 per violation | In force; first case under investigation |
| Washington | SB 5152 (2023) and successors | Disclosure duty + “forged digital likeness” offense | In force |
| Maryland | SB 141 (2026) | Year-round election deepfake ban (unique nationally); up to $25,000 and 20 years | Took effect in 2026 |
| Tennessee | HB 1513 (eff. July 1, 2026) | Disclosure in political ads within 120 days of an election; Class C misdemeanor | In force — but covers state races only |
| Hawaii | Act 191 (2024) | Time-limited ban | Ruled unconstitutional, Jan 2026 (First and Fourteenth Amendments) |
| Montana | SB 25 (2025) | Disclosure within 60 days of an election; graduated penalties | Being challenged by a PAC; no ruling yet |
Full list of all 31 states: see the Public Citizen tracker (data as of June 15, 2026).
The Line Courts Have Drawn: Bans Fail, Labels Survive
Line up the challenged laws and the pattern is unmistakable:
- What fell were broad bans: California’s AB 2839 (struck in its entirety, First Amendment), Hawaii’s Act 191 (First and Fourteenth Amendments), and California’s AB 2655 (platform takedown duty, preempted by Section 230).
- What survives are narrow disclosure rules: the labeling duties in California’s AB 2355, Michigan, Oregon, and Washington remain untouched.
- Minnesota’s narrow ban is holding on — but that was a procedural win, and the fight on the merits continues; Montana’s disclosure law is under challenge.
Legislators have clearly read the rulings: nearly every new law in 2025–2026 takes the disclosure route. “You can fake it, but you have to admit it’s fake” — that is where American election deepfake law stands.
Sexual Deepfake Laws: 46 States
As of January 2026, 46 states have laws targeting sexual deepfakes (Ballotpedia), 18 of which cover both child sexual abuse material and adult non-consensual images. Since the TAKE IT DOWN Act took effect, state laws have become the enhancement layer above the federal floor — the federal law sets the criminal baseline and platform duties, and states stack on top:
- Michigan (signed Aug 2025): creating non-consensual sexual deepfakes is now a crime; aggravated distribution is a Class F felony, up to 3 years.
- New Jersey (April 2025): deceptive deepfakes are a third-degree crime, up to 5 years and $30,000 — a law driven by Francesca Mani, the student at the center of the Westfield High School deepfake incident.
- New York (2023): distributing deepfake intimate images is a Class A misdemeanor, with a civil claim on top.
- Texas (2023→2025): SB 1361 covers adult images; three 2025 bills expanded it — dropping the “intent to deceive” element and adding a threat-to-distribute offense.
The New Frontier: AI-CSAM Offenses and Platform Liability
The third block has formed fastest: as of July 2026, 45 states have dedicated criminal laws for AI-generated child sexual abuse material (Enough Abuse tally; still missing: Alaska, Colorado, Massachusetts, Ohio, Vermont, and D.C.). Arkansas’s Act 977 (2025) is the model: AI content is criminal if it is “indistinguishable from a real child” — no need to prove it depicts any actual child.
Criminal convictions have arrived: a Pennsylvania man received 14 years and 7 months for deepfake CSAM; an Alaska case brought 70 months. But every one of those punishments landed on users. What about the platforms that supply the tools?
On July 24, 2026, an Arkansas family sued xAI, alleging Grok was used by a photographer — who already faces 200 state charges — to generate CSAM of minors. It is one of the first family lawsuits in the country aimed squarely at an AI platform, and the family’s lawyers say more are weeks away. Criminal law has reached the people; platform liability is still blank space — and this lawsuit is an attempt to write the first line in it.
The 2026 Midterms: The First Big Test of 31 State Laws
Enough theory — here is combat. The November 2026 midterms are the first nationwide stress test of America’s election deepfake laws, and primary season has already supplied plenty of material. Four states, four scripts, and both parties on the field.
Kentucky: Do AI Ads Work? The Loser Said So Himself
The KY-4 congressional primary was the most expensive House primary in U.S. history (about $25 million) — and the fiercest AI ad war. A super PAC opposing Thomas Massie (MAGA Kentucky) ran AI-generated images of Massie “holding hands” and “dining” with AOC and Ilhan Omar — with satire and AI disclaimers attached. Rep. Marjorie Taylor Greene publicly claimed the ads violated the TAKE IT DOWN Act; legal experts widely doubt that, since the law covers only intimate images and the ads carried disclaimers. The pro-Massie PAC (Kentucky 4th) answered with an unlabeled AI video: opponent Gallrein retreating on a battlefield while an AI-generated Trump charges forward, rifle in hand. On May 19, Massie lost to the Trump-endorsed Gallrein — and delivered the primary’s most quoted line: those AI ads “really worked on the boomers.”
Michigan: One State, Three Playbooks
Senate Majority Leader Nesbitt (running for governor) posted an AI video of Governor Whitmer standing in front of a tractor bearing down on her — his campaign called it “parody”; critics called it an incitement to political violence. U.S. Senate candidate Rogers posted an unlabeled “Hulk” AI video — legal, thanks to the state law’s satire exemption. And state senate candidate Tunney took an opponent to court over a deepfake video, moving in April 2026 for a default judgment of more than $157,000, with state police investigating in parallel; the case is unresolved. The reflection from Tsernoglou, co-author of Michigan’s disclosure law, says it all: “When we drafted this, we never imagined AI ads being used to incite political violence.”
Oregon: The Enforcement Machine Turns for the First Time
In the OR-5 primary, candidate Lockwood posted multiple unlabeled synthetic videos mocking U.S. Rep. Bynum — including scenes of her handing a locker-room key to a drag queen and attending a “satanic ritual” with Governor Kotek. Secretary of State Tobias Read opened an investigation under SB 1571 — the first formal enforcement of Oregon’s disclosure law, with fines of up to $10,000 per violation; no conclusion as of late July. Watch this one: it will deliver the first answer to “what does breaking a disclosure law actually cost?”
Tennessee: A Loophole Three Weeks In
The TN-6 story from the update log deserves the full picture. HB 1513 took effect July 1, requiring AI disclosure in political ads; three weeks later, in the 6th District’s Republican primary, Hilleary and Garrett traded undisclosed AI attack ads (an opponent’s campaign bus driving into a swamp; an opponent shaking hands with a drag queen and posing with Pelosi). Both ads were perfectly legal — because HB 1513 covers only state-level races, and congressional races belong to federal law, which has no disclosure requirement. However dense state law gets, federal campaign ads remain no-man’s-land. That is the real regulatory map of 2026.
And that is just the sample. Also in progress: Arizona legislator Chaplik sued a super PAC over undisclosed AI billboard ads (filed July 9); an AI video from Spencer Pratt supporters in the LA mayor’s race passed 5 million views; the NRSC released an 85-second AI deepfake of Texas Senate candidate Talarico with a disclaimer so small critics called it meaningless; and an all-AI campaign ad ran in St. Charles County, Missouri — a state with no disclosure law at all.
The big picture? Public Citizen’s assessment: most campaign AI content so far is either labeled or obviously satirical — the doomsday scenario has not arrived. But researchers caution that there is little empirical evidence labels actually change voters’ judgment. And every landmark enforcement case — the Oregon investigation, Michigan’s Tunney case, the Arizona lawsuit, Montana’s constitutional challenge — remains unresolved as of today. The laws exist; the teeth haven’t grown in yet. November will tell.
Deepfake Laws Around the World
Beyond the U.S., 2026 is the year deepfake rules land in force worldwide. Three major jurisdictions, three routes: the EU regulates through labels, South Korea through heavy criminal penalties, China through the full content pipeline.
European Union: Labeling Becomes Mandatory on August 2
The AI Act’s Article 50 transparency obligations take effect August 2, 2026: deploying deepfakes in the EU — with or without intent to deceive — requires clear disclosure that the content is AI-generated or manipulated. Art and satire get lighter treatment (disclosure in a way that doesn’t spoil the work), but no exemption. Fines run up to €15 million or 3% of global annual turnover. One detail that is easy to misread: the machine-readable watermarking duty for providers is postponed to December 2 for systems already on the market before August 2 — so what actually bites in August is mainly the deployer’s disclosure duty. The European Commission adopted its official Article 50 guidelines on July 20; and the parallel Digital Omnibus adds one hard rule: AI systems whose purpose is generating CSAM or non-consensual intimate images are banned from the market and from use.
South Korea: The World’s Heaviest Criminal Package
South Korea’s sexual deepfake laws are the strictest anywhere: creation or distribution carries up to 7 years; mere possession or viewing is a crime too, up to 3 years; distribution for profit starts at a 3-year minimum (in force since September 2024). For elections, deepfake campaign content that is hard to distinguish from reality is banned outright for the 90 days before a vote, punishable by up to 7 years — and ahead of the June 2026 local elections, the government deployed an AI detection model to enforce it. On January 22, 2026, the AI Framework Act took effect, mandating labels on AI content that is hard to tell from reality (with a one-year grace period on fines) — the world’s second comprehensive AI law after the EU’s, and the first to be fully in force.
China: First Mover, Regulating the Entire Pipeline
China’s Deep Synthesis Provisions, in force since January 2023, were the world’s earliest deepfake-specific rules. Since September 2025, the AI-Generated Content Labeling Measures and a companion mandatory national standard require AI content to carry both an explicit label (visible to users) and an implicit one (metadata); platforms must preserve labels through distribution, downloads, and re-uploads, and the rules assert extraterritorial reach. In April 2026, regulators released draft rules on “digital humans” for public comment: using someone’s likeness or voice would require explicit, revocable consent, and digital humans would need prominent labeling throughout (note: still a draft, not yet in force).
Also Worth Remembering
- Canada: Bill C-16 took effect July 18 — non-consensual sexual deepfakes are now criminal, threatening to distribute carries up to 10 years, and platforms must remove within 48 hours
- United Kingdom: distribution was already criminal (Online Safety Act); since February 6, 2026, creating (or commissioning) sexual deepfakes of adults is a crime too
- India: since February 20, platforms must prominently label “synthetically generated information,” and the takedown window for unlawful content shrinks to roughly 3 hours — or the platform loses its safe harbor
- Australia: federal law banned distribution in 2024; New South Wales criminalized creating, altering, and distributing from February 2026
- Brazil: psychological violence against women using AI/deepfakes is now an aggravating factor, adding up to half the sentence
- Denmark: a proposal to give every person copyright-style protection over their own likeness (50 years after death, satire exempt) — still pending
- Japan: no dedicated law yet; the Justice Ministry set up a deepfake study group in April 2026
Across jurisdictions, the rules are converging: disclosure and labeling are becoming the global consensus — U.S. state laws, the EU, South Korea, China, and India all land in the same place. For creators and platforms, that is the most certain compliance baseline there is.
Are Deepfakes Illegal? A Plain-English FAQ
Are deepfakes illegal in the US?
It depends on the content, not the technology. Non-consensual intimate images: a federal crime (TAKE IT DOWN Act), plus laws in 46 states. Election deception: regulated in 31 states. AI-generated sexual content involving minors: a federal felony, with dedicated offenses in 45 states. Satire, parody, art, and consensual content remain legal. A deepfake isn’t illegal by itself — what you do with it is what decides.
Is it illegal to make a deepfake of someone without consent?
It depends on the content. Intimate imagery: illegal — publishing it is a federal crime. Commercial use (using someone’s face or voice to sell things): most states’ right of publicity lets them sue, and the NO FAKES Act would upgrade that into a federal right. Ordinary jokes and memes: mostly legal, though defamation and state personality-rights laws can still apply.
Can you go to jail for making deepfakes?
Yes — people already have. A Pennsylvania man received 14 years and 7 months for deepfake CSAM; the Maryland athletic director who faked a principal’s voice got 4 months; an Ohio man became the first TAKE IT DOWN Act conviction in April 2026 (sentencing pending). The federal baseline: up to 2 years for publishing non-consensual intimate images, 3 years when minors are involved — and many states go higher.
Are deepfakes protected by free speech?
Partly. Satire and parody are protected by the First Amendment — courts struck down California’s and Hawaii’s broad bans in 2025–2026 precisely because they swept in satire. But courts have upheld labeling requirements: you have the right to create, not the right to pass fiction off as fact. Non-consensual intimate images and child sexual content have never been protected speech.
What should I do if someone made a deepfake of me?
In this order:
- Report it to the platform — under the TAKE IT DOWN Act, platforms must remove non-consensual intimate images and known copies within 48 hours.
- If the platform doesn’t act, file a complaint with the FTC at TakeItDown.ftc.gov.
- If you are a minor (or the image dates from when you were), use NCMEC’s Take It Down tool — the hash is generated on your device and the original image never leaves your phone; adults can use StopNCII.org, which works the same way.
- If there is extortion, fraud, or interstate harm, report it to the FBI’s IC3 — in 2025, the FBI broke out AI complaints for the first time: 22,364 reports, roughly $893 million in losses.
- Preserve evidence (screenshots, URLs, timestamps) before deleting anything; most states also offer civil claims, and once the DEFIANCE Act passes, federal statutory damages start at $150,000.
Is it legal to use deepfake apps at all?
The tools are legal; the boundary is how you use them. Legal: your own likeness, consenting adults, labeled satire and art, licensed commercial work like film dubbing and face replacement. Illegal: any non-consensual intimate imagery, any sexualized content involving minors, impersonating candidates before an election, impersonating anyone to commit fraud. In short: consent plus disclosure are the two safest lines you can draw.
Do deepfakes have to be labeled?
It depends on where you are and what you post. United States: most of the 31 state election laws require exactly that — labels — but there is no general federal labeling mandate. European Union: from August 2, all deepfakes must be disclosed. China: explicit plus hidden labels since September 2025. South Korea: the AI Framework Act requires labeling content that is hard to tell from reality. India: prominent labels required since February 2026. The trend points one way: labeling duties will only multiply.
Key Dates
- 2019
Texas SB 751: the first election deepfake law in the U.S.
- Jan 2023
China’s Deep Synthesis Provisions take effect: the world’s first dedicated rules
- Feb 2024
FCC rules AI-cloned voices are “artificial voices”; robocalls require prior consent
- Sep 2024
South Korea’s sexual deepfake criminal law takes effect (possession and viewing included)
- May 19, 2025
TAKE IT DOWN Act signed; criminal provisions effective immediately
- Aug 29, 2025
California’s AB 2839 and AB 2655 struck down on the same day
- Sep 1, 2025
China’s mandatory AI-content labeling takes effect
- Jan 13, 2026
DEFIANCE Act passes the Senate unanimously, again
- Jan 22, 2026
South Korea’s AI Framework Act takes effect: the first comprehensive AI law fully in force
- Apr 7, 2026
First TAKE IT DOWN conviction (guilty plea; sentencing pending)
- May 19, 2026
TAKE IT DOWN 48-hour platform takedown duty takes effect
- Jun 18, 2026
NO FAKES Act clears Senate Judiciary unanimously
- Jul 1, 2026
Tennessee HB 1513 takes effect (loophole surfaces three weeks later)
- Jul 18, 2026
Canada’s C-16 takes effect
- Aug 2, 2026
EU AI Act Article 50 and California SB 942 take effect the same day
- Nov 3, 2026
U.S. midterms: the big test for 31 state laws
Sources
Officialcongress.gov bill pages (S.146 · S.1837 · S.4591), CRS report LSB11314, FTC.gov enforcement releases and warning letters, DOJ press releases, FCC 24-59, the European Commission’s AI Act hub, and China’s CAC Labeling Measures.
TrackersPublic Citizen (state election laws, as of June 15, 2026), Ballotpedia (sexual deepfake laws, as of January 9, 2026), Enough Abuse (AI-CSAM, as of July 16, 2026), MultiState (AI bill totals).
ReportingNPR, PBS, Axios, Michigan Advance, Nashville Banner, Phoenix New Times, KATV, JURIST, and the Arizona Capitol Times, plus the per-entry links in the update log above.