Welcome back to The Hidden Layer. I’m Ian Krietzberg.

In today’s issue, a close look at a 4-year-old startup—with plenty of A-list fans in Hollywood—that’s determined to tackle the scourge of deepfakes and A.I. impersonators. I have some exclusive details about their expansion plans and new advisory board, and their surge in sign-ups following the release of OpenAI’s Sora 2.

But first: news and notes on a super PAC’s war against a New York state assemblyman; a new policy initiative from some of the major A.I. players to support deregulation in “blue and purple states”; and new research on the public health risks associated with data center expansion.

Also discussed in this issue: Keanu Reeves, Brad Pitt, Loti, Trump, Luke Arrigoni, Alex Bores, Greg Brockman, Koustubh Bagchi, the Chamber of Progress, Craig Kallman, Greg Smith, Michael Sugar, David Kaefer, Ian Calderon, James Feldman, Karl Austen, Shaolei Ren, Jensen Huang, and many more…

Let’s get into it…

Three Things You Should Know…

A.I. takes on the Upper East Side: On Monday, Leading the Future, a super PAC loaded with $100 million in funding from the likes of Andreessen Horowitz and OpenAI executive Greg Brockman, announced a multimillion-dollar campaign against New York State Assemblyman Alex Bores, who is running to succeed Rep. Jerry Nadler as congressman for most of the Upper East Side, Upper West Side, and Midtown Manhattan. Bores, of course, became persona non grata in Silicon Valley after authoring New York’s newly passed Responsible A.I. Safety and Education Act—a bill that Leading the Future described as “exactly the type of ideological and politically motivated legislation that would handcuff not only New York’s, but the entire country’s, ability to lead on A.I. jobs and innovation.”

When I gave Bores a call, he said he wasn’t surprised to be targeted. “They were signaling for a while that they wanted to spend a lot of money against anyone who wanted to put forward reasonable A.I. regulation,” he told me, noting that he worked with all the major A.I. labs to incorporate their feedback on the bill. Alas, Bores said, a subset of the tech industry just “doesn’t believe in the legitimacy of regulation, period.”
Meanwhile, in the states…: The Chamber of Progress, a pro-tech political advocacy organization whose partners include Nvidia, OpenAI, Apple, Google, and Amazon, launched a new policy initiative called the State A.I. Leadership Project, focused on educating policymakers in “blue and purple states” about “how A.I. can make government more effective, drive economic growth, and expand opportunity for residents.”

In a statement, Koustubh Bagchi, the organization’s policy chief, argued that blue states are authoring way more A.I.-related laws than red states, to the detriment of the latter. “Congress should pass federal A.I. legislation, but states aren’t waiting in the meantime, and too many proposals are driven by A.I. doomers that don’t reflect voter optimism,” he wrote. “We’ll be pushing state legislators to focus on what A.I. can do for their constituents and consumers instead of what to fear from it.”

A vast, bipartisan majority of American voters has supported strong A.I. regulation for years now. In September, Gallup found that 80 percent of U.S. adults think the government should prioritize A.I. safety and data security, even if it means slowing down on capability development.

A MESSAGE FROM OUR SPONSOR

McKinsey’s latest State of AI in 2025 report makes one thing clear: the next wave of AI advantage will be won by leaders, not algorithms.

High-performing organizations are three times more likely to have leaders who own and model AI usage, weaving it into strategy, culture, and execution.

The question is no longer if AI delivers value—it’s how leaders will capture what’s next.

Read the report at McKinsey.com/StateofAI

The public health cost of A.I.: A new analysis out of UC Riverside found that the electricity consumed by data centers in California nearly doubled between 2019 and 2023—and, not surprisingly, the associated carbon emissions almost doubled as well. As researcher Shaolei Ren first wrote last year, the industry is currently in the midst of a vicious cycle wherein data center energy demands are fueling rising carbon emissions and general air pollution, either through on-site gas generators or off-site electricity generation.

The new report projects that this added pollution could lead to thousands more cases of asthma by 2028. Already, the economic cost of data center–related pollution rose from around $44 million in 2019 to $155 million in 2023, and could reach as much as $266 million by 2028. “California can pioneer a sustainable model for data center growth,” Ren said in a statement. “But it requires continued commitment from California policymakers and regulators to keep lowering power grid emissions.”

Deal of the Week: Anthropic’s $15B Day

On the same day that Google launched Gemini 3—a model that, if benchmarks are to be believed, far outpaces the competition—Microsoft and Nvidia jointly announced that they were getting into bed with Anthropic. And Anthropic, whose major investors so far have included Amazon and Google, announced that it will adopt Nvidia’s architecture and scale its models using Microsoft’s Azure cloud service—a $30 billion compute commitment. In exchange for that newfound business, Nvidia will invest $10 billion back into Anthropic, while Microsoft will contribute $5 billion.

In a decidedly weird 10-minute video heralding the partnership, Microsoft C.E.O. Satya Nadella said that “we’re increasingly going to be customers of each other.” Nvidia C.E.O. Jensen Huang said that “the thing that’s really great is they’re gonna need a lot more Azure compute, and they’re gonna need a lot more G.P.U.s.” CNBC reported that the deal values Anthropic at $350 billion.

Nadella made sure to note that OpenAI remains a “critical partner,” but clearly Microsoft is diversifying its A.I. bets—especially as this investment cycle enters a more circular, fragile phase. Indeed, it isn’t clear how Anthropic—which loses billions every year—plans to meet that $30 billion compute commitment, though a $15 billion injection from the sellers of said compute will certainly help. (Anthropic didn’t respond to a request for comment.)

And now for the main event…

Hollywood’s A.I. Doppelgänger Hunters

Among the infinite challenges introduced by the A.I. boom, deepfakes and digital impersonators are among the most pernicious. Loti, an A.I. startup catering to the entertainment industry, hopes to solve the problem at scale.

Ian Krietzberg

While “finding you everywhere” might seem like a pretty Big Brother-ish slogan for an A.I. company, it’s actually the heart of Loti’s sales pitch. The 4-year-old startup helps customers like Keanu Reeves identify and scrub their digital imitators from the internet, using a facial recognition algorithm that can identify individuals in videos and images based on 15 percent or less of their face. The tech was developed to address the revenge porn crisis, but in 2024, following a $7 million fundraising round, Luke Arrigoni, Loti’s founder and C.E.O., expanded the company’s offerings to include voice recognition, shifted its clientele to celebrities and the entertainment industry, and enlarged its scope to encompass the entire internet.

In retrospect, the pivot was timely. As the A.I. boom continues to introduce novel challenges—from energy consumption to market bubbles to hyper-targeted voice scams—the public is perhaps most attuned to the issue of deepfakes. The problem is particularly vexing for the entertainment industry, where personal brands and reputation management are the coin of the realm. And the risks keep evolving as digital fraud becomes more sophisticated. In one viral case earlier this year, a French woman was scammed out of $850,000 by an A.I. Brad Pitt. “We shut down the deepfakes. We shut down explicit materials, false endorsements, etcetera,” Arrigoni said. “We find all those and shut them down.”

When we first spoke in July, Arrigoni told me that Loti was experiencing “mass adoption” across the entertainment industry, largely fueled by word of mouth. The company not only identifies images of clients in unauthorized places, but also issues automated, real-time takedown requests to the platforms hosting the content. Arrigoni told me that most platforms automatically remove offending content upon receiving a request from Loti—and that Loti’s now 60-person team can get content removed from an adult site within half an hour.

A MESSAGE FROM OUR SPONSOR

McKinsey’s latest State of AI in 2025 report makes one thing clear: the next wave of AI advantage will be won by leaders, not algorithms.

High-performing organizations are three times more likely to have leaders who own and model AI usage, weaving it into strategy, culture, and execution.

The question is no longer if AI delivers value—it’s how leaders will capture what’s next.

Read the report at McKinsey.com/StateofAI

Loti’s business surged again last month after OpenAI released Sora 2—an event that drove a 30-fold increase in subscriptions, Arrigoni said, with interest coming from groups ranging from talent agencies to law firms. “All of our clients were concerned that they’d show up in Sora,” he told me. “In the last few weeks, we’ve been downloading more content. So we’re able to give full vision to the entertainment industry on what’s being generated, who’s being generated, that kind of thing.”

Loti also has its sights set beyond Hollywood: Just today, the company announced the formation of an advisory board to help it strategically scale to new industries. Members of the board include Craig Kallman, chair and C.E.O. of Atlantic Records; Greg Smith, chair of American Airlines; Michael Sugar, founder and C.E.O. of Sugar23; David Kaefer, a former Spotify executive; Ian Calderon, a gubernatorial candidate in California; and James Feldman and Karl Austen, both managing partners at major Hollywood law firms. In an emailed statement, Austen gushed that “Loti has built a way to protect against this new A.I.-enabled exploitation with cutting-edge technology of its own, and will be an important piece of the puzzle in fighting emerging threats to digital rights in this new reality.”

Eating the Internet

Loti’s system is built around 40 different machine learning models, each designed to perform a specific function within its portfolio of services. Unsurprisingly, infrastructure costs aren’t cheap, coming in at around $1 million a month, but the system is optimized for precision and efficiency: Rather than deploy recognition algorithms across every possible frame, for example, Loti’s engineers designed a system that tells its models where to look when scanning content. The company runs this process hundreds of millions of times per day—which is why, Arrigoni told me, they are able to “eat the internet for only $1 million a month.”

At the moment, Arrigoni said, Loti’s facial recognition algorithm is something like 99.8 percent accurate—far better than humans at identifying faces based on partial exposure—and its speaking and singing voice recognition algorithms are around 97 percent accurate, a number he wants to improve with scale. The company’s free consumer offering is currently in beta mode and already fully subscribed, with the waitlist growing every day. “A lot of people are going to be going through crises, and we’re trying to make them feel like they have someone on their side,” Arrigoni said.

Going forward, Loti sees music and voice recognition as their next big frontier. The company’s algorithms can already isolate and analyze voices placed over instrumentals—an initial hurdle the technology had to clear. The bigger challenge, Arrigoni told me, is sussing out whether a customer’s voice is authorized for use in a given piece of content. “We are really trying to protect creative control and creative usage of who you are,” he said. “Even if you made a cartoon or music, there’s some creative expression. If you own that, Loti should be the platform that goes out and makes sure it’s still under your control.” Getting that context right, while keeping infrastructure costs low, will require a lot of technological sophistication.

Of course, the problem of deepfakes has also become an extremely active area for legislation, with a flurry of state and federal bills aiming to protect kids, artists, and celebrities alike. A few of those, like Tennessee’s Elvis Act, have already been signed into law, and the Take It Down Act, which prohibits the publication of nonconsensual intimate imagery, was one of the first pieces of federal A.I.-related legislation to be signed into law by Trump. A.I.-powered nudify apps, meanwhile, have become the subject of new laws in New Jersey and legal skirmishes in California, battlegrounds that A.I. companies and social media platforms alike are watching closely.

And yet, so far, the platforms are moving carefully when it comes to self-policing. As legislators and regulators race to keep pace with a technology that has blurred the line between reality and fiction, there’s been a push for social media companies to embrace watermarking, for instance, to clearly label synthetic content. Even then, there’s the underlying problem of digital doppelgängers showing up where they don’t belong. Loti doesn’t see that market going away anytime soon.

That’s all for today. I’ll see you on Thursday.

Ian

Join Emmy Award-winning journalist Peter Hamby, along with the team of expert journalists at Puck, as they let you in on the conversations insiders are having across the four corners of power in America: Wall Street, Washington, Silicon Valley, and Hollywood. Presented in partnership with Audacy, new episodes publish daily, Monday through Friday.

A professional-grade rundown on the business of sports from John Ourand, the industry’s preeminent journalist, covering the leagues, players, agencies, media deals, and the egos fueling it all.

The Phoebe Philo Paradox

LAUREN SHERMAN

WBD Deadline Drama

MATTHEW BELLONI

Glen Powell vs. Tom Cruise

SCOTT MENDELSON

Need help? Review our FAQ page or contact us for assistance. For brand partnerships, email ads@puck.news.

You received this email because you signed up to receive emails from Puck, or as part of your Puck account associated with {{customer.email}}. To stop receiving this newsletter and/or manage all your email preferences, click here.

Puck is published by Heat Media LLC. 107 Greenwich St., New York, NY 10006