Protesters march through midtown Manhattan to call attention to the effects of wealth inequality on: climate change, the Trump administration's foreign and domestic policies, and other social and political issues in New York on September 20, 2025. (Photo by Matthew Rodier/Sipa USA via AP Images)

Casey Michel on Breaking the Tech Oligarchy's Grip

Chief District Judge Bryan Biedscheid delivers closing remarks after the jury reached a verdict against Meta at New Mexico's First Judicial District Court, Tuesday, March 24, 2026, in Santa Fe, N.M. (Nathan Burton/Santa Fe New Mexican via AP, Pool)

Why New Mexico v. Meta Matters

hand putting ballot paper in box, ballot paper, ballot box, binary code, election, voting, voting ballot, democracy, politics, politics and government, technology

AI Didn't Create the Election Infrastructure Problem. It Just Made It Urgent.

Image of map and computer language over binary codes on server racks in server room.

Global Digital Policy Roundup: July 2026

Attorney General Raul Torrez, is joined by a group of mothers advocating for a change to social media, at a news conference outside First District Court, in Santa Fe, N.M., Monday, May 4, 2026. (Eddie Moore/The Albuquerque Journal via AP, Pool)

New Mexico Court Orders Meta to Pay to Establish $567 Million Fund to Abate Harms to Youth

Senate Commerce Committee Chairman Ted Cruz (R-Texas) and Ranking Member Maria Cantwell (D-Wash.) confer during a markup hearing on August 5. Yuqing Liu/Tech Policy Press.

Senate Commerce Committee Chairman Ted Cruz (R-Texas) and Ranking Member Maria Cantwell (D-Wash.) confer during a 

 on August 5. Yuqing Liu/Tech Policy Press.

Senate Committee Advances Four Bills Aimed at Protecting Children Online

Minister of the Supreme Electoral Court Kassio Nunes Marques attends the trial of former President Jair Bolsonaro at the Supreme Court in Brasilia, Brazil, Friday, June 30, 2023. (AP Photo/Eraldo Peres)

Brazil's Electoral Court Wants Platform Compliance Plans. Can It Enforce Them?

The OpenAI logo appears on a smartphone screen with the Microsoft logo in the background. (Photo by Samuel Boivin/NurPhoto via AP)

Why Merger Law Misses Frontier AI's Ownership Structure

Participants, including Nvidia Corp. CEO Jensen Huang (front, 2nd from R), Japanese Economy, Trade and Industry Minister Ryosei Akazawa (front, C) and Noetra Corp. President Hironobu Tamba (front, 2nd from L), pose for a photo during an event hosted by the Ministry of Economy, Trade and Industry in Tokyo on July 16, 2026. (Kyodo via AP Images)

What Sovereign AI Budgets Actually Buy

European Commission President Ursula von der Leyen, center, speaks during a report on children's safety online with co-chairs of the Special Panel Maria Melchior, right, and Jörg M. Fegert, left, at EU headquarters in Brussels, Monday, July 13, 2026. (AP Photo/Marius Burgelman)

Europe's Summer of Digital Services Act Enforcement Targets Platform Design

A pixelated black-and-white portrait of Ada Lovelace where the arrangement of pixels forms intricate borders and repeating patterns. These designs resemble the structure and layout of GPU microchip circuits, blending her historical contributions with modern computational technology.

Lovelace GPU by Hanna Barakat & Cambridge Diversity Fund / 

How AI Hype Helps Render Women Invisible

Police artificial intelligence technology wearing blue vest outfit isolated digital collage. Cop with vintage tv for head with AI text.

Will AI Sycophancy Contaminate Law Enforcement?

Aerial view of the work on White House Helipad on August 1, 2026 in Washington, D.C. Credit: mpi34/MediaPunch /IPX

Five Questions the US Government Should Answer About Its Secretive Frontier AI Framework

On the left, there is an image of a tree on a snowy, rocky landscape, with light shining in from the background. On the right, there is the same landscape, but it is now reproduced as grey and silver cubes in the shape of a tree and technical terrain, backgrounded by the light.

Linus Zoll & Google DeepMind / Better Images of AI / Generative Image models / CC-BY 4.0

The Model You Audit Is Not the Model You Ship

Sen. Josh Hawley (R-Mo.) and Sen. Richard Blumenthal (D-Conn.) speak before a Senate Judiciary Subcommittee hearing on artificial intelligence at the US Capitol, in Washington, D.C., on Tuesday, July 25, 2023. The two senators are among those sponsoring the GUARD Act. (Graeme Sloan/Sipa via AP Images)

AI Is Not Social Media. New Child Safety Laws Should Reflect That.

European Commission President Ursula von der Leyen speaks during a report on children's safety online at EU headquarters in Brussels, Monday, July 13, 2026. (AP Photo/Marius Burgelman)

Does the EU's Special Panel Succeed in Charting a Third Path for Child Safety?

Earth movers prepare a site for a 2.5 million square foot AI data center Tuesday, March 24, 2026, in Independence, Mo. (AP Photo/Charlie Riedel)

Big Tech’s PR Playbook for Overcoming Data Center Opponents

OpenAI founder and CEO Sam Altman is seen with aides in an elevator after a meeting with Sen. Ted Cruz (R-Texas) on Capitol Hill on July 29, 2026. (Francis Chung/POLITICO via AP Images)

July 2026 US Tech Policy Roundup

Google CEO Sundar Pichai talks before a dinner with President Donald Trump and Japan's Prime Minister Sanae Takaichi in the State Dining Room of the White House, Thursday, March 19, 2026, in Washington. (AP Photo/Julia Demaree Nikhinson)

Google's AI Governance Plan Draws the Boundaries of What Counts as Harm

A view of the Berlaymont European Commission building in Brussels, Belgium on July 20, 2026. (Photo by Jakub Porzycki/NurPhoto via AP)

EU AI Act Transparency Rules Are Now In Effect. Was It A Missed Opportunity?

Audio of this conversation is available via your favorite podcast service.

The rise of the tech oligarchy threatens democracy in the United States and abroad. In his new book, 

, Casey Michel—author, journalist, and director of the 

 with the Human Rights Foundation—tells the story of how a small set of extremely wealthy individuals claimed the commanding heights of American government and is making the world safe for authoritarians, and what will be necessary to break their grip. 

United States of Oligarchy
How America's Wealthiest Ally with Dictators, Weaken the U.S., and Destroy Democracy
Author: Casey Michel

On Thursday, the First Judicial District Court in Santa Fe, New Mexico 

State of New Mexico v. Meta Platforms Inc. 

The court found Meta created a public nuisance by contributing to the youth mental health crisis and by facilitating child sexual exploitation. It ordered a range of remedies, including that Meta pay $567 million to create an abatement fund to finance various measures intended to address the harms to the public.

“In sum, the Court finds that New Mexico is in the midst of a teen mental health crisis affecting public health and public safety in and throughout the state, and that Meta’s platforms are a significant contributing cause to the crisis,” wrote Chief Judge Bryan Biedscheid in the decision.

The decision follows the second phase in 

State of New Mexico v. Meta Platforms Inc., 

which consisted of a bench trial. Its central question was whether Meta's platforms amounted to a public nuisance in New Mexico, and, if the court found that they did, what remedies would be needed to address it. In March, a Santa Fe jury found Meta liable for violations of New Mexico’s Unfair Practices Act, awarding 

 following the nearly seven-week first phase of the trial. 

. “We remain confident in our record of protecting teens online and will continue to defend ourselves against claims that misrepresent the facts,” a Meta spokesperson said.

If established, the requirement for Meta to pay $567 million into an abatement fund would be in addition to the civil penalties, according to today’s decision, and would be put toward different categories of activity, including treatment, awareness and prevention, and ongoing monitoring.

Table titled "Abatement Fund" from paragraph 121 of a court order, breaking down the $567,000,000 abatement fund Meta is ordered to pay, allocated across five categories:

Awareness and Prevention: $33,000,000.00
Screening and Assessment: $90,000,000.00
Referral, Linkage, and Coordination: $15,000,000.00
Treatment: $420,000,000.00
Implementation, Continuous Quality Improvement, and Evaluation: $9,000,000.00
Total: $567,000,000.00

A breakdown of the categories of abatement measures stipulated in the decision. 

New Mexico Attorney General Raúl Torrez 

, alleging the company made false public statements about the safety of its platforms while knowing internally that its products facilitated child sexual exploitation. The court 

In today’s decision, the court asserted that “Section 230 does not preclude the State’s public nuisance claim,” but the decision attempted to thread the needle on issues that the court determined might have run “afoul” of the statute, or of the First Amendment, such as issuing remedies around any particular product feature. 

Nonetheless, regulation of industry-wide features that contribute to problematic use, such as infinite scroll and autoplay, require legislative or executive branch action; if the Court were to order changes to such features, it would risk competitively damaging Meta’s platforms to an unreasonable extent, and possibly running afoul of the First Amendment and Section 230 because of the direct effect those features have on content presentation.

 to force Meta to remove such features. Nevertheless, in a 

, Torrez called the decision another landmark victory.

“This case has always been about protecting children, standing up for families, and making sure that one of the world’s largest technology companies cannot profit from practices that endanger young people without consequence,” said Torrez.

Nathaniel Lubin served as an expert witness in State of New Mexico v. Meta Inc.

State of New Mexico v. Meta Platforms, Inc.

 should prove a landmark decision in the battle to make Big Tech accountable for its products' harms: for the first time, a social media platform was found to have caused a “public nuisance” due in part to the mental health epidemic among children on the platform. While headline fines ($942 million across the two phases) will drive attention, the most important aspect of the verdict is the finding that technology platforms may be held legally responsible for the 

 of their products on critical populations of users.

A focus on these types of effects belies the popular narrative of social media as a persistent, small, drag on the productivity and wellbeing of all users. Instead, like prescription opioids, the trial revealed why this view of small effects on typical users is wrong and obscures platforms’ much more catastrophic effects on identifiable minorities of users, a reality that may well be echoed in future litigation.

As one of the state’s expert witnesses in the trial, one particular 

 from 2017 illustrates the key dynamic. Amid a discussion among employees about internal research on excessive use of Instagram, the photo-sharing app that Facebook acquired in 2012, Head of Instagram Adam Mosseri, wrote: “1.3 percent of people in the US spending 30-plus hours a week [on Instagram] sounds like a lot.” The staff member who conducted the research responded with a remark saying that it was a judgment call as to whether that was a big enough group to justify a response.

It is hard to understand asking whether a million people is a big enough group to deserve help. Yet that’s just what the decision requires: that treatment of such issues receive more than a perfunctory response.

The government’s case was vindicated in March when jurors in the first phase of the trial 

, on all counts, that the company had misrepresented the safety of its social-media products. This week’s ruling built on the jury’s finding by focusing on the “public nuisance” created by exposure to social media, similar to the legal responses to products like tobacco.

Judge Bryan Biedscheid’s decision is nuanced, and gives important procedural wins to Meta (which 

 it will appeal). But in establishing public nuisance for the first time – that “Meta’s platforms injure the public health, safety, and welfare in New Mexico” – he also ordered numerous binding requirements on Meta’s products and procedures. Specifically, in New Mexico he required that Meta:

Maintain private by default settings for all under-18 users, including preventing underage users from showing up in search results unless identified by specific username;

Prevent under-18 users from appearing in the recommendation algorithms of unconnected adults and prevent any unconnected adult from messaging an under-18 user;

Enforce a 1-strike policy for adult users who engage in child sexual exploitation, including all accounts related to such a user, and preventing the creation of future accounts;

Create a “clearly visible, forward-facing reporting mechanism” on every surface for unwanted sexual advances that receives a human review within 48 hours;

Prohibit “romantic or sexualized” interactions between children and AI chatbots;

 push notifications for under-18 users during both nighttime and schooltime hours, except for messaging from connected users and for security;

Implement mandatory time limits for under-18 users of not more than 90 hours cumulatively across Instagram and Facebook per month;

Show a warning banner to all adults at least once a month and to all under-18s at least once a week, and show screens to all under-18s highlighting safety tools and features daily.

Besides their direct effects (if upheld on appeal), these requirements are important for numerous reasons. Above all, they demonstrate the need to assess companies’ decisions based on 

 for the most affected users. The ruling publicly clarified that, for years, Meta’s relentless focus on maximizing product uptake came at the expense of users like the ones Mosseri referred to. And it offers a roadmap for both further litigation and future product design rules.

To date, critics of social media have overwhelmingly focused on the experience of 

 users, especially typical teenagers. Questions like “Has social media caused mental health harms to young people?” matter, and the case offered direct evidence that the answer to that question is “yes.” Yet answering that question requires assessing a range of effects across very large groups of people. Especially because independent researchers have lacked access to platforms’ internal data, companies have been able to dispute findings and deflect criticism.

user is often misleading. In the real world, platform engagement is extraordinarily uneven, meaning that a small share of users racks up a disproportionate share of platform activity. As a consequence, the most severe harms from use are also highly concentrated. Other data presented at trial showed that this pattern persisted: between 2020 and 2024, the most active 1 percent of New Mexico teenagers still spent more than 30 hours a week on Facebook and Instagram.

What is true for time usage is also true for platforms’ more acute problems, such as harassment and crimes like the distribution of child sexual-abuse material and human trafficking. The evidence presented showed that these patterns were consistent and that the actions the company took in response often were weaker than other available options.

When companies like Meta acknowledge abuse, they tend to describe such harms as unavoidable. Sentiments boil down to “we’re trying; the problem is really hard,” particularly familiar from Mark Zuckerberg’s congressional 

. Similar explanations were echoed in court. The executives’ arguments focus on procedural choices: look at all the accounts that were removed, they say, or see how we’re using machine-learning models to limit promotion of illegal activity.

Digital tools reproduce offline social problems, a challenging fact to manage. But the court heard tacit recognition that the measures supposed to control such problems were not scoped to solve them. For instance, the judge’s rule enforcing 1-strike policies may seem unnecessary given other descriptions of company behavior. Yet another internal email from 2020 read into 

Number one, perm block [is] actually just one year for terrorism, CEI [child exploitation imagery], CEI solicitation, revenge porn, content-based extortion, and sextortion. Temporary block [of] two weeks for nudity, bullying, NAZI, drug, gang, hate speech, cartel, graphic violence, and impersonation. Do not block for self-harm promotion, eating disorder promotion, copyright, trademark, counterfeit, spam, U13, sex offender, et cetera.

It turns out that enforcing 1-strike policies, including device blocking, may require an outside order.

While the enforcement set by the court will not solve every problem posed by social media in New Mexico, if scaled more broadly to other geographies and jurisdictions it will meaningfully change Meta’s management incentives. At the same time, the Judge’s decision not to engage with stronger rules related to issues like age assurance or features like recommendation algorithms, autoplay, and infinite scroll lend credence to the practical value for clarification by legislatures that such design rules are not protected by statute. (Though, notably, in other parts of his decision, the Judge forcefully makes the case that such rules related to content do not restrict abatement even as he subsequently declines to set restrictions on these features.) In Judge Biedscheid’s reading, 

 is a particular obstacle to orders related to age assurance, despite an FTC 

 from earlier this year on exactly this topic.

Given the radical way that machine learning is reconfiguring people’s lives, the public-interest case for influencing product incentives has never been more urgent. New technologies like large language models are black-box products far more complicated than social-media applications. And reasonable usage by average users, let alone by sophisticated engineers, is driving extraordinary payoffs and value. Yet, alongside these benefits, we must also anticipate abuse among a vulnerable minority of users. For instance, based on our knowledge of how social media operates, we can anticipate that people who spend the equivalent hours of a full-time job each week talking to chatbots are very likely to need extra protection, regardless of whether AI companies succeed in adding other safeguards.

, Rachel Carson argued that a chemical company is responsible for misrepresenting the safety of a pesticide when it knows the risk for the small group of people repeatedly exposed.

 can be a similar turning point in how we understand technology products. While people and companies should be free to pursue their wants and wishes, we cannot ignore the at-risk outliers. We should focus on them: not as vulnerable victims in the making, but as the canaries in the coal mine who warn the rest of us when the environment is becoming toxic.

 from a malicious actor. Last month, an OpenAI agent 

broke out of a sandboxed testing environment

 and hacked Hugging Face, among others. Researchers and trust and security experts 

. And a July report suggested European leaders anticipating 

 coming from the United States. All of these developments put together tell us one thing: cyberattacks on elections are only a matter of when, not if.

Ever since communities around the world have started using computer programs to generate voter rolls, machines to tally votes, and websites to publicize results, experts have feared interference that would undermine trust. They have been proven correct in many cases, such as the recently revealed 

 on its national elections. The private sector itself 

 significant attacks on elections, as well as political parties and candidates. Despite these warning signs, conversation about protecting elections is dominated by concerns about information integrity, disinformation campaigns, and deepfakes. These issues matter too. But they have absorbed most of the attention and most of the funding, while a more basic vulnerability — the technical fragility of the infrastructure elections actually run on — has gone comparatively under-resourced. Now

In April 2026, Anthropic disclosed a model, Claude Mythos Preview, reportedly capable of finding and exploiting software vulnerabilities more effectively than all but the most skilled human hackers. The company chose not to release it publicly, citing the damage a wide release could enable. Instead, it launched 

, giving roughly fifty organizations — including Amazon Web Services, Apple, Cisco, Google, JPMorgan Chase, Microsoft, and Nvidia — early access to find and patch vulnerabilities in software. That a leading AI lab judged this capability serious enough to withhold from general release was like a warning light. The more recent Hugging Face incident made it a blinking red alarm.The threat is no longer only that a malicious actor deliberately targets an election system; it's that election-adjacent infrastructure could be damaged or exposed by an AI system pursuing an unrelated goal that simply happened to be in its path.

No election management body or election technology vendor was invited to be part of Project Glasswing. Anthropic hasn't explained the omission, but the answer is obvious: elections aren't where the customers, liability, or revenue sit for a company deciding where to point a defensive resource like this, even though elections are central to the democratic systems that let these companies thrive. Nobody decided elections didn't matter; the issue simply was never going to land on a commercial roadmap on its own.

On top of that, who exactly were technology companies going to call if they did want to concern themselves with elections? Election management bodies are generally poorly positioned to address technology risks. Most operate with little or no dedicated cybersecurity budget or staff, leaning on other bureaucratic entities with competing priorities. Meanwhile, the cybersecurity industry effectively owns the threat signals they need to see, but election authorities have no standing channel into that knowledge. Engagement, when it happens, is ad hoc and election-by-election, usually concentrated in wealthier markets. Procurement rules compound the problem: even where officials and technologists want frank conversations about specific risks, the rules of engagement often prevent them from saying much.

Europe has started to take this seriously. The European Parliament's Special Committee on the European Democracy Shield has 

 that explicitly calls for electoral infrastructure to be designated as critical infrastructure, including through a revision of the EU's Resilience of Critical Entities Directive. That is a meaningful step: critical infrastructure status typically unlocks financial support, regulatory oversight, and coordinated response planning that election systems currently lack almost everywhere. The report is scheduled to go to a full plenary vote this fall. It will not, by itself, be binding on member states or the Commission, but it is a rare instance of a legislative body naming the structural problem out loud.

This is not a problem unique to Europe, and certainly Europe can’t solve it alone. The same structural gap is showing up wherever elections happen. In South Africa, ahead of the 2026 local government elections, the Independent Electoral Commission has itself 

flagged the scale of digital protection and cybersecurity risk

 the vote requires, prompting parliamentary scrutiny of whether the commission's safeguards are adequate. That is precisely the kind of national conversation a European model, done well, could help other countries have earlier and more systematically, rather than each election authority arriving at the same conclusion independently and under pressure. Having the conversation now will also increase trust in elections, rather than keeping it from public view and risking further distrust. Afterall, it is common sense that a malicious actor would want to attack an election, and common knowledge that election technology is overburdened and rarely cutting edge. Elections are a sitting duck for cyber attacks.

What the Democracy Shield report or the South African Election Commission cannot do is solve the trust and access problem between election authorities and the technology companies. Previous efforts, such as the US Cybersecurity and Infrastructure Security Agency (CISA) in the 

 to address technology in elections, serve as instructive models. That said, companies themselves have stated that they'd prefer a sustainable, neutral, non-commercial model that could reach all election bodies, especially those unable to afford enterprise-grade protection on their own. That requires an intermediary — one that election bodies will actually trust with sensitive information, and one that technology companies will engage with outside the pressure of an active election cycle. Civil society organizations are the most credible candidate for that role. They frequently identify electoral risks earlier than governments or companies do, through frontline monitoring work they already carry out.

As the European Democracy Shield process moves from report to implementation, civil society organizations have been designing this mechanism. The pieces exist in entities such as 

. Election officials convene regularly and have singled out the challenge and technology companies have readily participated. This is not a technical gap, it is an institutional one. The roadmap has been developed. It needs a greenlight and the funding to make it happen.

The threats driving this — whether a deliberate attack on voting infrastructure or an AI agent pursuing a goal that happens to run through it — are not going to wait for the plenary calendar. Europe has an opening now to lead the way for the world, and build the connective tissue between election authorities and the companies that hold the threat data, anchored by the civil society groups that already do this work. The cost of empowering such solutions is low, and certainly less costly than recovering the trust that will be lost in the aftermath of the election cyberattack that we know is coming.

Britain's Prime Minister Andy Burnham is shown the driver's area on a FirstGroup bus at the bus station during a visit to Bath, England, Wednesday July 22, 2026. (Toby Melville/Pool Photo via AP)

Burnham's First Week Redraws the Map of UK Tech Governance

Can the Pall Mall Process Curb Spyware Abuse?

A group of shareholders of Samsung Electronics Co. hold a rally criticizing a tentative wage agreement struck between the company and its labor union near the home of Samsung Electronics Chairman Lee Jae-yong in Seoul, South Korea, Thursday, May 21, 2026. The letters read "Unjustified transfer strike." (AP Photo/Ahn Young-joon)

Workers in Asia Are Fighting for Protections as AI Threatens Jobs

Tech Policy Press - Technology and Democracy

Tech Policy Press is tracking laws and regulations, along with government investigations and litigation, that will shape the rules and accountability for tech companies.

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe