Illustration of two men inflating a balloon labeled "tech"

Antitrust Enforcement Can Deflate the AI Bubble Before the Public Pays

Anthropic CEO Dario Amodei (not pictured) and Chief Product Officer Mike Krieger unveil Claude 4 during the Code with Claude conference, announcing both Opus 4 and Sonnet 4 AI models to developers and industry leaders on Thursday, May 22, 2025 in San Francisco. (Don Feria/AP Content Services for Anthropic)

What AI Companies Can Learn from the Oversight Board

A data center and its backup diesel generators built by the Markley Group loom over a ballpark and residential neighborhood in Lowell, Mass., on June 30, 2026. (AP Photo/Matt O'Brien)

Governments Can Advance 'Greener AI' Through the Power of Procurement

PICTURE POSED BY A MODEL File photo dated 21/08/14 of a child using an Apple iPhone smartphone. Schools cannot address issues with children's smartphone use and screen time alone, headteachers have warned. Delegates at the annual conference in Belfast for the NAHT school leaders' union voted in favour of a motion calling for national guidance on online safety to help families establish boundaries for their children with screens. Issue date: Saturday May 2, 2026. 84481834 (Press Association via AP Images)

Europe Can Protect Children Online Without Surveillance or Age Bans

A woman is passing by the logo of Europese Commissie - Commission Europeenne - European Commission, at the Berlaymont building, the building of the European Commission Brussels, Friday 17 April 2026. BELGA PHOTO MARIUS BURGELMAN (Photo by Marius Burgelman/Belga/Sipa USA)(Sipa via AP Images)

(Photo by Marius Burgelman/Belga/Sipa USA)(Sipa via AP Images)

Europe's W Social Bet Tests its Vision of Digital Sovereignty

In this photo illustration, a close-up of a printed circuit board (PCB) is visible, while the ASML logo appears on a screen in the background. ASML supplies lithography systems used in semiconductor manufacturing. (Photo by Nikos Pekiaridis/NurPhoto via AP)

(Photo by Nikos Pekiaridis/NurPhoto via AP)

Why ASML's Semiconductor Monopoly Doesn't Give Europe Strategic Control

The image shows OpenAI CEO Sam Altman sitting in a dimly lit conference environment, illuminated by soft blue lighting that creates a shadowy, mysterious ambiance. He is dressed in a dark suit with a light-colored shirt, and they are gesturing with one hand as if making a point or addressing an audience. 

Less Apocalyptic Rhetoric Can Help Mitigate Anti-Tech Violence

The Supreme Court’s Decision in Trump v. Slaughter is Misguided, But Changes Little

Capitol Building, Montgomery, Alabama. Photo from the Carol M. Highsmith Archive collection at the Library of Congress via Wikimedia Commons.

Capitol Building, Montgomery, Alabama. Photo from the 

Where State AI Legislation Stands Half Way Into 2026

Southampton, UK—June 2, 2026: Riot police push forward with shields as violent clashes erupt as protesters gather to express their anger at the murder of Henry Nowak. (Photo by Christopher Walls / Sipa via AP Images)

New UK Prime Minister Should Elevate Defending Democracy Taskforce to Tackle Bots and Protect Trusted News

Tech, Power, and the Struggle for American Democracy

Photo by Zackary Drucker and Alyza Enriquez, 

GLAAD Maps Where AI Fails LGBTQ People and How to Fix It

The U.S. Supreme Court is seen Monday, June 29, 2026, in Washington. (AP Photo/Mariam Zuhaib)

Who’s Regulating Police Technology? It’s Not the Courts.

The image is a very detailed, black-and-white sketch-like illustration featuring a complex scene of interconnected figures and technology. The artwork portrays various individuals in different environments to represent the relationship between technology and humans. 

In the foreground, multiple people are surrounded by computer screens filled with data visualizations, charts, and technical information. A woman seated in an armchair appears deep in thought, surrounded by data-filled monitors. Beside her, a man leans over, using a tablet to assist with their inspection of a plant or tree. In the centre, a figure holds a large frame or screen displaying anatomical illustrations, representing the use of AI to analyse medical imagery. To the left, another person is intently observing a computer screen, while a second figure nearby is deeply immersed in analysing data. A woman dominates the right side of the composition, gazing upwards as if in contemplation or envisioning something beyond the immediate scene. The background features more people, including a family holding hands, and other abstract representations of data.

AI is Everywhere by Ariyana Ahmad & The Bigger Picture / 

Why We Need a 'Truth Campaign' for the AI Era

Protesters against the Trump administration gather in front of the US Capitol in Washington, D.C., on April 5, 2025. (The Yomiuri Shimbun via AP Images )

A Judicial Wake-Up Call on Government by AI

A sign opposed to data centers is held during a protest against AI data centers in Vancouver, British Columbia, Canada, on Saturday, June 27, 2026. (Ethan Cairns/Bloomberg via Getty Images)

No Data Centers in Anyone's Backyard

World Summit on the Information Society (WSIS)+20 High-Level Event 2025 by “WSIS” by UN Geneva, CC BY-NC-ND 2.0

World Summit on the Information Society (WSIS)+20 High-Level Event 2025 by “

As the UN Launches its Global Dialogue on AI Governance, WSIS Offers Critical Lessons

America 250 rally on the National Mall for the Great American State Fair. Wednesday June 24, 2026. (Official White House photo by Patrick B. Ruddy)

America 250 rally on the National Mall for the Great American State Fair. Wednesday June 24, 2026. (Official White House 

June 2026 US Tech Policy Roundup

An electoral official sits at a polling station before polls opened for the presidential runoff election in Santiago, Chile, Sunday, Dec. 14, 2025. (AP Photo/Natacha Pisarenko)

Countering Authoritarian Disinformation Requires More than Platform Governance

Members of the UN Scientific Panel on AI. Source

Members of the UN Scientific Panel on AI. 

AI Will Affect Everyone's Future. Its Governance Can't Belong to a Few.

A woman is passing by the logo of the European Commission at the Berlaymont building in Brussels, Belgium, on Friday, 17 April 2026. (Photo by Marius Burgelman/Belga/Sipa USA)(Sipa via AP Images)

 that it was joining W Social, a new European social media platform built on the AT Protocol, the move attracted attention as an apparent effort to align its communications strategy with its broader digital sovereignty agenda.

The announcement followed the Commission’s release of its 

, a set of legislative proposals to boost the bloc’s autonomy over its digital infrastructure, which includes support for decentralized open networks. Yet the move to W Social has also prompted criticism. Although W Social does not appear to meet the EU’s own Open Source Strategy, senior EU leaders — including European Commission President 

The episode raises broader questions about what "digital sovereignty" means in practice. In a 

 on tech sovereignty, the European Commission outlines that digital sovereignty is understood as the EU's capacity to exercise its independence in the digital domain, ensuring the EU retains its “ability to decide, invest, and innovate according to European values of democracy, openness, and the rule of law.” Acknowledging that Europe is highly dependent on US technology, the need for greater control over its digital infrastructure has been highlighted as a top priority for the EU.

The Commission’s Open Source Strategy outlines the ambition not only for Europe to wean itself off US Big Tech but also to pursue a different approach to European tech adoption, recognizing the benefits of a more open and distributed approach to tech adoption that avoids the pitfalls of vendor lock-in. However, its actions signal a pursuit of tech sovereignty which focuses only on where infrastructure is located, ignoring the need to also 

pursue sovereignty through openness, interoperability and user choice

Launched at the 2026 World Economic Forum

 as a for-profit European social platform governed by European law, hosted on European infrastructure, and "designed for conversations among humans around the world." The platform positions itself as a European alternative to X. Its principal distinguishing features are identity verification intended to reduce bots and its use of the AT Protocol, the decentralized networking protocol developed by Bluesky.

The company’s CEO and co-founder is Professor Anna Zeiter, previously a professor in Data Protection Law at the University of Bern with experience working at companies like Flo and eBay. Ingmar Rentzhog is the co-founder and sits on the board of W Social and is a co-founder and CEO of the media firm 

On the surface, its uptake by the European Commission is a positive sign that decision makers in Europe are signaling interest in moving away from platforms based in the US. Although the Commission already maintains a 

 — including operating its own server — it continues to rely heavily on commercial platforms such as X and LinkedIn for public communication. However, artist and open-network advocate 

 that W Social had removed its public GitHub repository, 

 from developers and open-source advocates, who viewed the move as inconsistent with the collaborative ethos of the AT Protocol ecosystem.

Understanding the criticism requires understanding the goals of the AT Protocol itself. Although often associated with the 

 by the Bluesky team as a shared infrastructure that allows multiple applications to interoperate rather than tying users to a single platform. This simplifies app creation for developers, who can build on top of an open protocol and plug straight into a network. Rather than competing against apps to concentrate all users on their individual platform, users can mix and match apps and components of apps without having to rebuild their accounts and networks of followers and friends.

 of the protocol is to facilitate the creation of an ecosystem of apps that users can easily move between or use in combination, which is diametrically opposed to the dominant model of centralized platforms like Facebook and Twitter, which are designed to lock in users. Going beyond enabling competing apps within the ecosystem, the protocol

 “to enable decentralization by having multiple interoperable providers for every part of the system.” This means users control their feeds and how they are presented, and can pick their own content moderation services within the apps on the protocol. Though technical commentary has questioned the 

 to how decentralized from Bluesky the AT Protocol ecosystem can become, the AT Protocol was designed with “credible exit” in mind to avoid the pitfalls of remaining trapped within “

With openness at the heart of the project, major apps like the Bluesky social network and 

, a social networking organization that began as a custom feed on Bluesky centering black voices, have both open sourced their code, which is 

. Though it is not mandatory to open source its code, it is in keeping with the philosophy behind the protocol to do so. Indeed, openness has meant that developers are able to learn from each other.

For instance, Blacksky, which has grown from a custom feed to 

 a fully open source and decentralized stack of technologies for communities to build their own social networks, 

 to develop a similar European social networking infrastructure based on the AT Protocol. Sherif Elsayed-Ali, a co-lead for Eurosky, 

 that the team can build new apps quickly “because we’re building on open source foundations and years of work by others. We contribute back by open sourcing our work so others can use it. Open source is an incredible force multiplier.” This illustrates why it is significant that W Social has chosen to close its codebase; it directly contradicts the project of the AT Protocol. Because the company has chosen to close its GitHub repository, other developers cannot build on their work or build interoperable services with W Social.

Part of W Social’s pitch is to build a space where only verified humans can participate. For users who sign up for W Social or migrate their existing Bluesky accounts to W Social, their account is still viewable to those using other AT Protocol apps, but W Social 

 identity verification for posting, commenting, messaging, and interacting with users on the W Social platform and the feed is only viewable from the W Social app. To verify, users must use a separate app, W Identity, to scan either a passport or a national identity card, prompting 

However, this approach undermines the interoperability that the decentralization movement aims to achieve. The goal is to let users access networks through the apps or communities of their choice, mix and match feeds and algorithmic ranking, and customize content moderation to meet their needs. By building on the open protocol, W Social 

 the benefit of directly plugging into the entire existing network comprising over 

 users. However, critics say it is also attempting to draw users into a closed ecosystem. Elsayed-Ali 

 that while there are cases for closed community spaces to be built, public apps that position themselves as a public square “should not be gated.”

 that while repositories were previously publicly accessible, they were made private while “we reviewed security, privacy, licensing, intellectual property, documentation and commercial considerations.” They have said that they may again publish “selected components, improvements, integrations, bug fixes and technical documentation” in the future upon review. They reiterated a commitment to supporting open standards, interoperability and user choice, as well as contributing “responsibly” to the AT Protocol ecosystem, while balancing a need to build their own “secure, sustainable and distinctive service.”

However, by closing its codebase and attempting to shy away from the interoperability that the AT Protocol ecosystem was built to facilitate, it moves closer toward the model of the centralized platform, locking in users, denying user choice and control, the very issues that adopting open source and decentralized networks are supposed to fix. The mere fact that its servers are based in Europe is not sovereignty if users are denied control.

Online commentary on the Commission’s move to W Social has questioned why the agency has backed a closed source platform mere weeks after publishing its 

. The strategy outlines a set of recommendations by the Commission to support the adoption and development of open source technology, arguing that it can be a successful strategy for Europe to maintain control over its digital infrastructure. At the heart of the strategy is the pursuit of a digital ecosystem that is “interoperable, portable and based on the widespread adoption of open standards – to ensure that public and private users can choose, switch and scale technological solutions without prohibitive costs.” The open strategy includes tech throughout the entire stack, from cloud computing to social networking.

The strategy makes clear the benefits of open source, pointing out that it promotes secure and auditable digital infrastructure, enabling independent inspection and peer review, while also boosting competitiveness. Similar to the motivations behind the AT Protocol, the strategy points out that open source lowers the barriers to market entry, and reduces costs by fostering “collaborative innovation by enabling communities and companies to jointly develop, adapt and secure technologies.”

The strategy lays out objectives not only to “strengthen and promote a vibrant open source ecosystem” but also to promote and adopt “open and interoperable digital ecosystems for public administrations, including EU institutions.” Recognizing the Commission “as a major public administration and policymaker,” the strategy states that the administration will use and develop open source broadly, “contributing to a European open and sovereign digital ecosystem.”

Mastodon’s Executive Director, Felix Hlatky, told Tech Policy Press that “we believe the European Commission is making thoughtful decisions,” citing the Commission’s Mastodon server and the Tech Sovereignty Strategy as demonstrating “that they understand sovereignty isn’t just built at the server level but throughout the entire technology stack.”

However, Sebastian Vogelsang, a co-lead at Eurosky, 

 to the Commission’s adoption of a closed-source app with "surprise," given the aims outlined in the strategy to support open source. He also 

 that in a recent workshop on the future of social media, researchers from the Commission’s Joint Research Center had “presented recent work highlighting the importance of open social media infrastructure and specifically referenced initiatives such as Eurosky”, further adding confusion to the decision to migrate to W Social.

Nicholas Gates, Senior Policy Advisor at OpenForum Europe, told Tech Policy Press that “this decision doesn’t reflect well on the Open Source Strategy.” “There is a need to have the right principles beyond just seeing that this is built on an open protocol. It also shows some of the contradictions inherent in implementing a strategy like this and reveals some of the underlying friction in implementing an Open Source Strategy in such a large bureaucracy.” Gates, however, does not think that this decision dooms the strategy as a whole, but it shows that implementation is going to be uneven at times.

Jean Cattan, the Future of Technology Institute’s Of Counsel, warned that the European Commission’s policy choice could risk favoring closed solutions and signals “a narrow pursuit of digital power.” He warned that the Commission should protect its aspirations of an open, distributed, and shared digital economy, arguing that “we should redirect momentum toward genuine market openness rather than a race for power that would run counter to the founding principles at stake.” Aline Blankertz, co-founder of the digital-policy collective Structural Integrity, similarly added that the Commission’s pick of W Social “illustrates the Commission's interpretation of sovereignty: European companies should imitate the principles of US companies instead of building on principles that structurally limit dependencies.”

This may be the first stress test for the AT Protocol, one that tests how well centralization resistance is built into the protocol’s structure. Bluesky’s 

, “the organization is a future adversary" articulates the project of the AT Protocol to ensure that users “

 pack their bags (take all of their data and relationships with them) and move to a different server.” Infrastructurally, users cannot be locked in. W Social may become the protocol's first significant test of that premise.

The question is not simply whether users can leave the platform, but whether Europe's digital sovereignty agenda will ultimately favor infrastructure that keeps exit possible or platforms that seek to recreate familiar forms of enclosure under European ownership. Will the Commission proceed to pursue an open source strategy that recognizes the benefits of a distributed approach to tech adoption and avoids the pitfalls of vendor lock-in, or one that settles for sovereignty based on server location?

request for comment by Tech Policy Press.

(Photo by Nikos Pekiaridis/NurPhoto via AP)


Calls for European technological sovereignty are growing louder. In early June, the European Commission unveiled a 

, including a Chips Act 2.0, aimed at reducing the bloc's dependence on foreign technologies.

Days later, the limits of that ambition became clear. The US government ordered AI company Anthropic to 

 to its most advanced models for users outside the United States before 

. For nearly three weeks, a single directive from the United States cut Europeans off from Anthropic's most capable AI models before Washington lifted the restrictions after negotiations with the company. The episode illustrated how quickly access to a frontier technology can become subject to US export controls.

But the incident revealed something bigger. Even when critical technologies are developed or used outside the United States, Washington often retains the ability to determine who can access them. Nowhere is that tension more visible than in Europe's most strategically important technology company: ASML.

The Dutch company spun out of Philips in 1984 and builds the photolithography machines that print the circuitry for computer chips. For the most advanced chips, ASML is the world's only manufacturer of extreme ultraviolet (EUV) lithography systems, the machines required to produce them. Every leading-edge processor produced for companies such as Apple, Nvidia and AMD depends on ASML equipment somewhere in the manufacturing process. In June, it became the first European company ever to pass a market value of 

, making it the most valuable listed company in the continent's history.

To many policymakers, that monopoly looks like Europe's greatest source of geopolitical leverage. If the world cannot build advanced chips without ASML, then surely Europe has a powerful bargaining chip. The question is whether Europe can actually exercise that leverage. That position has led some policymakers to view ASML as a potential geopolitical chokepoint. Economists describe such positions as technological chokepoints: single points in a supply chain through which entire industries must pass. 

, a widely circulated scenario by AI researchers examining Europe’s technological future, treats ASML as one of the continent’s few remaining sources of strategic leverage.

Niclas Poitiers and Tillman Schenk from the Brussels think tank Bruegel describe this dynamic as 

. But that argument rests on a key assumption — that Europe can act as a unified geopolitical actor in export controls.

Let’s start with the market. ASML's sales are almost entirely in Asia and the United States, with Taiwan's TSMC as its largest customer. As ASML chief executive Christophe Fouquet recently 

less than one percent of the company's more than 32 billion euros in revenue came from the European region in 2025. Europe has fallen behind, he said, and encouraged more investment in the AI and semiconductor ecosystems.

Now consider the technology. ASML draws on a network of thousands of suppliers across dozens of countries for its products. Its optics come from Germany's Zeiss. Critical precision components are sourced from suppliers in the United States, Japan and Taiwan. And the light source at the heart of every machine is developed and built in San Diego, California, by an ASML division that grew out of the American firm Cymer. The supply chain — and the technology base — are anything but exclusively European, and that matters. 

 (FDPR) allows Washington to control products made outside the United States if they rely on American technology or software. It is one of the main mechanisms through which US export controls extend beyond its borders. The monopoly is Dutch in corporate terms but embedded in a broader transatlantic technology stack.

Then there is the law. It would not be the first time the Netherlands has restrained ASML, but it has always done so under American pressure. In 2019, after a visit by then Prime Minister Mark Rutte to the White House, The Hague 

 for the most advanced machines bound for China. In 2023, further restrictions 

 on additional equipment, a significant blow to ASML, since China accounted for roughly a third of its revenue. In both cases, Dutch policy moved in step with US strategic priorities.

Some, like Niclas Poitiers and Tillman Schenk, 

 that if Washington depends on Dutch cooperation to restrict ASML’s advanced lithography from China, then that dependence becomes a form of European leverage. The reasoning is sound, but it conflates "Europe" with the Netherlands. Export control is a national competence under EU law, and Washington prefers to deal with individual capitals at a time. The apparent European chokepoint is therefore exercised at the national level.

But the dependence is not symmetrical. The Netherlands has far fewer substitutes for US markets, components, and regulatory alignment than the United States has for Dutch consent. Washington also retains unilateral reach through the Foreign Direct Product Rule, a tool The Hague cannot match. The result is asymmetry rather than shared leverage.

What began as diplomatic pressure now edges toward legislation. Before the US Congress sits the 

, a bipartisan bill introduced in April that would prohibit the sale and servicing of advanced machines to Chinese chipmakers, a double commercial blow for ASML. The bill would give the Netherlands and Japan 150 days to align their export controls with the American ones. If they fail to, it directs Washington to extend the Foreign Direct Product Rule to any machine containing even a single American component or piece of software. For ASML, that means nearly everything it builds. Whether that extension ever materializes is uncertain. But the threat alone is real.

This month, the pressure arrived directly when US Commerce Secretary Howard Lutnick 

 that Washington believes one of its extreme ultraviolet machines may have reached China, in breach of the very ban that has kept EUV out of the country from the start. ASML rejected the claim. It said it has never shipped an EUV system or a purpose-built component to China, and that it can account for every installed machine through remote telemetry. With the MATCH Act still under discussion, Dutch officials have increasingly engaged Washington directly. Rather than apply pressure of its own, The Hague sent a trade delegation to 

. It was an illustration of where leverage currently sits.

Europe does hold a unique instrument in ASML, a real chokepoint in the advanced chip industry. But it cannot wield it freely, because the market, the technology, and the legal weight all lie on the other side of the Atlantic. Anyone who thinks this is a theoretical scenario need only look at the Anthropic episode. The one country that can turn the screw on Europe's most valuable company is not the Netherlands. It is the United States.

For years, Europe has argued about how best to keep children safe online. The fight has come down to two options: scan their messages or lock them out of many platforms until they're older. There is a third way, one that protects children without watching them. Yet Europe has never defined, in any testable way, what a "safe" platform for children actually looks like.

first proposed Child Sexual Abuse Regulation

 ("Chat Control" to its critics) would have obliged platforms to reach inside private messages, issuing detection orders for child sexual abuse material (CSAM), and for the solicitation of children (i.e. grooming). To reach inside end-to-end encrypted messages, detection has to happen on the device itself, before the message is encrypted, a technique known as client-side scanning, and security researchers have said repeatedly that 

there is no way to build a backdoor that only good people can walk through

compromise texts going into the final trilogue

 had narrowed detection back toward hash-matching (comparing files against a database of abuse images) for known material on unencrypted servers. Mandatory scanning has been blocked, for now, and there’s still no agreement on what should replace it. Meanwhile, the separate 

temporary derogation allowed platforms to scan voluntarily expired in April 2026

. Though that expiry may not stick. The council has since moved to reinstate it, and Parliament will 

, this time under different rules that will make it easier to be approved.

Denmark is legislating a social-media age limit for under-15s

. The intent is right. But the mechanism is blunt. We already have a preview of how it goes: when the UK switched on age-verification rules for adult content in 2025, 

— one provider reported a 1,800 percent jump in the first month. If a determined 13-year-old can defeat an age gate with a free app, a ban that depends on the same technology can be evaded the same way.

Both fixes assume someone must watch the child

Both options are usually fought as separate battles: one over how to tackle abuse, the other over whether young children belong on social platforms at all. They are not the same problem, but they rest on the same hidden assumption: that protecting a child requires someone — an adult, a server, a regulator, a parent — to 

 what the child is doing. Scanning watches everything. Bans check who you are before they let you in. Neither asks whether protection could happen without surveillance at all.

It can. The technology to detect grooming as a conversation unfolds can run on the child's own device — flagging a risk pattern locally and surfacing a calm, age-appropriate intervention in the moment, with no message content, metadata or signal ever leaving the phone. There is no database to breach, no cache of data to subpoena or repurpose and no encryption to break because nothing is transmitted.

objections that security researchers raise against client-side scanning 

are that whoever controls the on-device match list — the set of things each phone is told to look for — controls what every phone searches for, and that the moment a detection signal leaves the device, you have enabled surveillance on the phone.

A third way answers the second objection outright: the signal stays on the device and nothing is transmitted onward. It changes the shape of the first, too — the model looks for conversational risk rather than matching against a centrally controlled list of targets. But it does not make that first concern vanish: any detection model has to be trained and, over time, updated, and whoever controls those updates still has a say in what the phone looks for. The answer to that cannot be "trust the vendor." Instead, the model, and every update to it, must be auditable and certified against an open standard, so that the thing running on a child's phone can be independently checked and cannot be quietly re-pointed at something else.

The obvious objection is that a system that flags a risk, but does not report it, cannot deliver safety. But that assumes that the only protective act is escalation to an adult. Behavioral evidence points the other way: a timely, well-designed prompt at the moment of risk can change what people, including adolescents, do next. The research behind the anti-cyberbullying tool ReThink found that when teens were prompted to reconsider before sending a hurtful message, 

adolescents changed their minds 93 percent of the time and decided not to post it

. While ReThink measures how to get a child to reconsider a message they are about to send, not warning them they are being targeted by an adult, the transferable finding is the mechanism: a prompt delivered at the decisive moment reliably changes what the child does next.

There are two different functions at play, and it helps to keep them apart. Detection-and-reporting is a platform-level legal duty: where the law requires it, the platform detects the material and reports it to the authorities. On-device intervention is not that. It does not detect an illegal artifact and route it to anyone; it recognizes a risk pattern in a live conversation and surfaces a prompt to the child. The two can coexist: any legal duty to detect and report CSAM continues to sit at the platform level, where it already does. On-device intervention adds a different layer — the live conversation before the harm is done. The grooming-specific evidence base is still young, which is precisely why it should be built in the open and measured against a standard, not asserted behind a marketing claim.

There is another issue that on-device prompts resolve: proposed solutions that would escalate potential solicitation to a parent or guardian assume that the adult in the loop is safe, present, attentive and on the child’s side. For many children, that assumption simply does not hold. Child sexual abuse is frequently committed by someone the child knows, 

, one of the most comprehensive studies to date, found that, in the countries where this is measured, around one in 13 children experience sexual abuse by a family member and that relatives accounted for more than 55 percent of assessed CSAM, with fathers estimated to have produced over 900,000 images globally in 2024. Even when the adult is not a threat, they may be absent, unaware, or ill-equipped to act. A safety model that escalates to a parent assumes a parent who is watching and capable. This is not to say an on-device prompt reaches every child — contact abuse in the home is a different problem — but it is why 'escalate to the nearest adult' cannot be the default safety architecture.

When we build systems like this, we need shared, testable definitions of what a well-designed intervention actually is. It should not be left to each vendor's judgment. At a minimum, it means an intervention that changes what the child does in the moment, which the evidence on timely prompts suggests it can. It must also do the opposite of harm: it must not frighten the child, must not teach them more about the threat than they need to know and must never make them feel they have done something wrong or that acting will get them or someone they love into trouble. That fear is precisely what keeps children silent — and an intervention that surfaces only to the child removes it. This also flips the usual problem of false positives. A scanning system must be near-perfect, because its mistakes fall on the accused; an on-device prompt can afford to err on the side of caution, because the worst case is a calm nudge shown to a child who was never at risk and can simply ignore it. This does not soften the privacy-versus-protection trade-off. It dissolves it.

No one has defined what makes a platform “safe” for children

It’s critical to note that a technique that has the potential to protect children without compromising encryption is not a guarantee. On its own, "we run an on-device safeguard" is just one more feature a company can claim protects children, and that no one can independently check. The platform could ship a weak version, or none, and from the outside, you could not tell. That is the crux of my argument’s second half: whether an on-device protection genuinely stays on the device, reports to no one, and actually works cannot rest on the platform's word. It has to be something an independent body can test against a shared standard and withdraw when it fails.

Which brings me to the deeper failure beneath Europe’s deadlock. Every instrument it has produced — the Digital Services Act, the AI Act, the GDPR, the UK's Age Appropriate Design Code, the new 

 that platforms protect children. Not one of them defines, in testable terms, what "protected" means. So when a platform says "we comply" or "we keep children safe," there is no shared, independent way to check whether that is true.

The cost of that absence is clearest in the one corner of this problem where we actually have hard numbers — the detection and reporting of CSAM. In 2023, Google

 filed roughly 1.47 million reports to the US National Center for Missing & Exploited Children; Apple filed 267

. That 5,000-fold gap is not a measure of how much abuse occurs on each service. It is a measure of how hard each chose to look. When detection is voluntary and undefined, "we found nothing" and "we never looked" are indistinguishable from the outside — and children pay the price.

We need an open standard of what a safe-for-children service actually does

That is why we need a standard: an open, auditable definition of what a safe-for-children service actually does, control by control, that an independent third party can certify against — the way we certify food hygiene, payment security, or building fire safety. Not a logo a company awards itself, but a verifiable claim someone else can check and, when it fails, withdraw.

Put the two together — detection that runs where the conversation is, and a standard that makes "safe" checkable, and you no longer have to choose between a child's safety and a child's privacy, because protection no longer requires surveillance. You no longer have to choose between a ban and a free-for-all, because "appropriate measures" stops being a phrase lawyers argue over and becomes a list an auditor can tick.

Europe is writing its next chapter on this right now. The 

 on a new action plan to protect children from crime has just closed, and negotiators have concluded what may be the final trilogue on the CSA Regulation, with a political deal expected within weeks if not days (and perhaps already struck by the time you read this). The instinct in both will be to reach, again, for the two familiar levers — scan harder or ban younger. But Europe should ask a different question. Not "how much should we surveil children to keep them safe?" and not "at what age should we lock them out?" — but rather "why are we still accepting *

* from the platforms, when we could simply require them to prove they are safe, without watching a single child?" The third way has been there the whole time. We have been too busy fighting over the first two to build it.

Disclosure: Custorian is non-commercial and takes no funding from the technology platforms its standards cover. It has a direct interest in the standards work and intervention research described herein.

Amid efforts by the Trump administration and federal lawmakers to stop or limit states from enacting new AI legislation, as of July 1, more than half of US states have together enacted more than 100 new AI laws this term.

Every December, the Center on Technology Policy at NYU 

 of what states did that year in AI and other areas of technology policy. But by July, most states have concluded their legislative sessions, allowing an initial look into the overall trajectory of the year. Below, I offer eight observations on the scope of state AI legislation so far in 2026.

1. States continue to actively pursue AI regulation across topics, despite continued efforts by the Trump Administration and federal lawmakers to limit state AI policymaking.

By our count, as of July 1, states have enacted 109 AI and 28 data center laws. This is slightly behind the pace of last year, when by the same date states had enacted 121 AI and 27 data center laws (by the end of 2025, states enacted 159 AI and 37 data center laws).

This activity has come despite a continued effort by federal officials to stop or limit state AI legislation. In 2025, a group of Republican senators 

 a moratorium on state AI laws as part of budget reconciliation. Although the Senate ultimately rejected that proposal, related preemption efforts continued to appear in other forms. In December, 2025, President 

 creating a new AI Litigation Task Force charged with challenging state AI laws that were not “minimally burdensome.” The order also directed the Department of Commerce to explore withholding federal broadband funding—known as 

—for states that enact "onerous" state AI laws.

2. Federal preemption efforts may have limited the number of states enacting AI regulation, even if they have not dramatically reduced the total number of enacted AI laws.

So far in 2026, 29 states have enacted AI legislation. Last year, 39 states had done so by this point in the year, and in total 46 enacted AI laws by the end of the year. Of the 21 that have not yet enacted AI laws this year, 10 still have ongoing legislative sessions. Some of these states, including California, are widely expected to enact AI laws before the end of the year. Even still, it appears fewer states are on track to enact AI laws in 2026.

That said, the states that have not yet enacted AI laws do not look obviously different from those that have. About 61% of states with Republican trifectas and 69% of those with Democrat trifectas have enacted AI legislation. As of July 1, the 

 that has enacted AI laws has $742.2 million in BEAD funding, while the average state that hasn't has $920.9 million in BEAD funding. Once California enacts an AI law, as it is widely expected to do, those averages will be much closer: $873.7 million for states with AI laws and vs $779.6 million for states without them. While federal pressure may have affected the overall legislative environment, we can not yet see a simple partisan or BEAD-driven pattern as to which states are most impacted.

3. Federal preemption efforts may have influenced the substance of state AI laws.

The AI and data center laws enacted in 2026 look notably different than those enacted in 2025. While there is always some year-to-year variation, there is reason to believe that federal preemption efforts helped shape the types of laws states enacted in 2026.

Specifically, at least two distinct trends are emerging: more cross-partisan alignment on AI legislation, and a shift across states toward child safety, data centers, and consumer protection.

 identified several categories of state AI laws that the federal government should not preempt, including state regulation addressing child safety or “AI compute and data center infrastructure.” So far, there has been substantial state legislative activity in both areas. Both topics would have likely attracted significant attention regardless of the EO. But by signaling that these categories were less likely to be targeted by federal regulators, the EO may have moderated the chilling effect that the EO may have otherwise had.

4. Compared to last year, there is more cross-partisan alignment on AI regulation.

From 2025 to 2026, there was no meaningful difference in the share of AI laws with a bipartisan mix of sponsors. In 2025, 32% of enacted AI laws had sponsors from both parties; in 2026, it is 31%. (Those percentages drop to 11% and 7% respectively when looking only at primary sponsors.) But, while there has not been more bipartisan legislation, we do see lawmakers in both parties increasingly addressing a similar set of topics. In other words, while some AI issues remain largely Republican or Democrat-led, there is a growing set of topics attracting cross-party interest.

For example, in 2025, two Democrat-controlled states, California and New York, enacted companion chatbot bills. This year, six states with Democrat trifectas and eight states with Republican trifectas enacted companion-chatbot or similar laws. A similar pattern is appearing in laws addressing AI and mental health, professional licensing, and nonconsensual intimate imagery.

5. Companion chatbots regulation has become a major initiative for both Republicans and Democrats.

As our last annual state technology policy report 

, AI companion chatbots were the most active area of state AI regulation in 2026. State legislators introduced more than 

. There are notable differences across these new laws, but most build on the two laws enacted in 

 in 2025. These laws require operators to include warnings that chatbots are not human and to address certain risks, including sexual content involving minors and self-harm.

Despite these commonalities, there are significant debates over several other provisions. Most notably, there has been a partisan split over age verification and parental supervision, with Republican-supported bills more likely to include stronger age verification and parental-oversight requirements.

 that remains under consideration in California would extend last year's companion chatbot law by requiring pre-deployment risk assessments. If enacted, it would represent a significant extension of the scope of companion chatbot legislation.

6. For the first time, both Republican and Democratic-led states enacted laws imposing more restrictions, oversight, or requirements on data centers.

For the last 15 years, state data center legislation has done little beyond granting data centers tax incentives. This began to change last year, with a 

 that created new rate classes or instituted permitting preform. So far this year, state legislators have already enacted 28 laws, many of which grant states more control of data center development and operation. Broadly, these laws fall into four categories:

Protecting ratepayers from energy cost increases.

 In addition to a handful of laws creating new rate classes for data centers or large load customers, several states enacted broader legislation to protect ratepayers from price increases. For example, Oklahoma legislators enacted the 

Data Center Customer Ratepayer Protection Act of 2026

, which requires electric companies to create new large load tariffs designed to ensure that large-load customers reimburse utilities for any costs “fairly allocated to them,” including “costs incurred to directly serve the customer that may remain unrecovered if the customer departs the system or materially reduces load.”

Establishing new oversight, audits, or disclosures. 

After several bills requiring new energy or water use audits 

 last year, legislators enacted several new requirements this year. Notably, these are not limited to Democrat-led states. 

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe