A bipartisan coalition with support from Hollywood power players and Prince Harry and Meghan Markle's Archewell Foundation is working to prepare U.S. voters for a possible deepfake onslaught as the campaign year goes into high gear.

Why it matters: With federal agencies and social media companies barely talking to each other about AI-driven misinformation threats, "this is a disaster waiting to happen — no one's doing the public inoculation," warns Miles Taylor, chief policy officer of The Future US, which is coordinating the campaign.

• Taylor is a former DHS chief of staff — and author of a celebrated op-ed critique of the Trump administration from an "anonymous" insider — who wrote a tell-all book.

The big picture: Last September, the Fifth Circuit Court of Appeals ruled that the White House, FBI and other federal officials likely violated the First Amendment by encouraging social media companies to crack down on COVID-19 misinformation.

• The Supreme Court, which is hearing a government appeal of that ruling, is skeptical that officials overstepped — but the case has already significantly reduced the contact between officials and big tech companies, including election-protection coordination.

What they're saying: "The government isn't talking to social media companies. Many of the social media companies don't want anything to do with the government — which means novel AI threats could get missed," Taylor says.

• Cybersecurity and Infrastructure Security Agency officials confirmed the lack of contact in a March reporter briefing.

Driving the news: The nonprofit, nonpartisan organization The Future US shared two prototype ads that the group is using to gin up interest among Hollywood screenwriters, ad executives and influencers for what it hopes will be a massive earned media campaign, supplemented by an initial $5 million of paid advertising in swing states.

• One ad shows a woman in Arizona picking up the phone on Election Day and hearing a voice on the other end that tells her to stay away from the polls because of threats from militant groups. The AI-generated caller, impersonating a poll worker, holds a "real" conversation with the voter.
• In another ad, a Florida man is enraged by viewing AI-generated security footage fraudulently depicting ballots being destroyed, and urges his friends to take action.

The AI + Election Security Coalition, operated by The Future US, is backed by figures ranging from former Google X head of experience Tom Chi to NYU AI expert Gary Marcus and former presidential candidate Andrew Yang, along with former members of Congress from both parties.

Reality check: It will take more than $5 million and celebrity stardust to prepare swing state voters for AI fraud, especially when many already distrust mainstream information sources.

Experts predict both malicious state actors and domestic political operatives will use deepfake technology and other media manipulation to confuse voters during the U.S. election campaign.

• Oren Etzioni, a former CEO at Allen Institute for AI, is offering free tools to journalists and fact-checkers for spotting AI disinformation through a new nonprofit, TrueMedia.org.

The other side: Google, Meta and OpenAI did not respond to a request for comment.

• All three have joined a 20-firm coalition pledged to combat deceptive AI election content.

Fun fact: Wondros, the production company led by Jesse Dylan, Bob Dylan's son, produced the ads, Taylor tells Axios.

What's next: The Future US will launch "satirical and comedic campaigns" later in the spring, per Taylor, focused on five crucial swing states: Florida, Pennsylvania, Arizona, Michigan and Wisconsin.

• Sub-groups that will be targeted include voters aged 65+ and those who might be targeted by voter suppression efforts.

The Taiwan Semiconductor Manufacturing Company will begin production of the world's most advanced semiconductors in Phoenix by the end of the decade, the U.S. Commerce Department announced today, reports Axios Phoenix's Jessica Boehm.

Why it matters: The plant, and another expected to open in 2028, will manufacture 2-nanometer chips — that's the most miniaturized circuitry to date, allowing for the most efficient and powerful semiconductors.

• The expansion will be aided by CHIPS and Science Act funding in Arizona, one of the closest election swing states.

What they're saying: "These are the chips that underpin all artificial intelligence," Commerce Secretary Gina Raimondo told reporters.

By the numbers: TSMC will receive up to $6.6 billion from the CHIPS Act, bringing its total Arizona investment to $65 billion — the largest foreign direct investment in a new project in U.S. history, Raimondo said.

• The third fabrication plant is expected to create about 6,000 tech jobs and 20,000 "high-paying construction jobs," she said.

Context: President Biden visited Arizona last month to announce $20 billion for Intel's U.S. expansion.

Rep. Cathy McMorris Rodgers (R-Wash.) and Sen. Maria Cantwell (D-Wash.) released the American Privacy Rights Act yesterday, aiming to set a national data privacy standard, report Axios Pro's Maria Curi and Ashley Gold.

Why it matters: Data is the core component of AI, and lack of a national standard in managing that data leaves AI users exposed and complicates life for AI companies complying with a patchwork of state and foreign rules.

What's inside: Consumers would be able to sue and seek monetary damages when companies violate their privacy rights.

• A new privacy office at the FTC and state attorneys general would also enforce the bill.

The federal standard would override state laws while preserving sector-specific state laws that protect financial, health, employee and educational data.

• The bill targets data brokers by establishing an FTC registry with a "do not collect" tool for people to use.

A version of this story was published first on Axios Pro. Unlock more news like this by talking to our sales team.

More than half of the 260 companies in Y Combinator's latest graduating group are building or using AI technology, reports Axios' Kia Kokalitcheva.

Why it matters: These startups are showcasing serious AI tech that can be built without raising hundreds of millions of dollars.

State of play: YC partner Jared Friedman tells Axios that GPUs, which have been in short supply, are now "more accessible" to startups, and that it's now possible for startups to train their own models because the large foundational model companies aren't using certain datasets.

• It creates opportunities for upstarts like new YC graduate Piramidal, which works with brain wave data.
• "No matter what happens with [OpenAI's] GPT-5, it's probably not going to do anything with brain waves because that's just not the data it's trained on," Friedman says.

Between the lines: The expansion of available tooling has also translated to more young entrepreneurs starting AI companies, according to Friedman.

• A Ph.D. in machine learning is no longer required, because a talented young software engineer can do it.

Zoom in: AI startups in this latest cohort have trained models focused on music (Sonauto), weather forecasting (Atmo), chemistry (Yoneda Labs) and proteins (Diffuse Bio); they also include various engineering applications, plus speech and images.

What we're watching: Friedman predicts in the next year or so we'll see more robotics startups in YC's program.

• AI leaders like OpenAI, Google and Meta have all "cut corners," broken rules or rewritten them in their relentless quest for more data to use to train AI models. (The New York Times)
• While some students are turning in essays written by AI, some teachers are using AI to grade them. (CNN)
• A popular AI homework-helper app called Gauth is owned by TikTok owner ByteDance. (Axios)

A palace fit for 19 feral cats.