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.

The Supreme Court is seen Tuesday, June 30, 2026, on Capitol Hill in Washington. (AP Photo/Jose Luis Magana)

Supreme Court Decimates Key Remedies for Tech Complicity in Human Rights Abuse

Social media logos are displayed on a smartphone alongside an under-16 age-restriction symbol in this photo illustration, as governments around the world introduce or consider bans and stricter age-verification rules to limit minors' access to social media platforms. (Photo by Jonathan Raa/NurPhoto via AP)


Can Regulators Actually Deliver a Presumptive Teen Social Media Ban?

The UN Global Dialogue on AI Governance Should Tackle the AI Language Gap

Hand holding a smartphone displaying the Anthropic "AI" logo, against a colorful collage background with illustrated icons including a code speech bubble, tangled scribble, geometric shapes, megaphone, and DNA helix.

Expressive Governance Is a First Amendment Threat Hiding in Plain Sight

United Nations Office Geneva located in the Palais des Nations building in Switzerland. Shutterstock

United Nations Office Geneva located in the Palais des Nations building in Switzerland. 

The Three Temptations Facing the UN's First Global AI Dialogue in Geneva

Broken Chair or Chaise Cassee is a wooden sculpture by the Swiss artist Daniel Berset and Louis Geneve outside the United Nations in Geneva, Switzerland. Shutterstock

 is a wooden sculpture by the Swiss artist Daniel Berset and Louis Geneve outside the United Nations in Geneva, Switzerland. 

In Geneva, the World Can Anchor AI Governance in Free Expression

The Capitol is seen in Washington, Monday, June 29, 2026. (AP Photo/J. Scott Applewhite)

Bipartisan Smorgasbord of Children’s Online Safety Legislation Passes the House

Collage-style illustration of two hands linking pinky fingers in a pinky promise against a cream paper-textured background. The hands are in black and white, with one wearing a red sleeve and the other a red sleeve with red nail polish. Decorative geometric shapes surround them: yellow circles, dark red rings, a grid pattern, parallel lines, and a cluster of navy dots.

Trust Issues Could Make or Break Agentic Commerce

The image shows a silhouette of a person, likely a woman, sitting on the floor in a dimly lit room. She is sitting near a large window with sunlight streaming in, creating a strong contrast between the light outside and the darker interior. The woman appears to be holding a phone or tablet, looking at the screen while leaning her arms on her knees. The overall mood of the image is reflective, perhaps suggesting a moment of contemplation or solitude. The setting seems to be indoors, with part of the room's furniture visible on the right-hand side.

Human Rights Experts Should Engage in Age Assurance Standards

The United Nations Office Geneva in the Palais des Nations. Shutterstock

The United Nations Office Geneva in the Palais des Nations. 

The World Doesn't Need Another UN AI Declaration. It Needs Architecture.

European officials release the “Tech Sovereignty Package” on June 3, 2026. Source

European officials release the “Tech Sovereignty Package” on June 3, 2026. 

Does Europe Really Have a Plan for Tech Sovereignty?

A neural network comes out of the top of an ivory tower, above a crowd of people's heads (shown in green to symbolise grass roots). Some of them are reaching up to try and take some control and pull the net down to them. Watercolour illustration.

How AI Keeps Europe Hooked on US Cloud

People hold signs in opposition of xAI's use of gas turbines at their facility during a meeting of the Memphis and Shelby County Air Pollution Control Board, Monday, Dec. 15, 2025, in Memphis, Tenn. (AP Photo/George Walker IV)

Inside SELC's Clean Air Case Against xAI in Memphis

The undying sun hangs in the sky, as people gather around signal towers, working through their digital devices.

Digital Nomads Across Time by Yutong Liu & Digit. 

Imagining Broadband Policy of, by, and for the People

An abstract background meant to depict pixel errors and glitches.

Nine Things Platforms Could Do Now to Help Blunt Political Violence in the United States

US Commerce Secretary Howard Lutnick listens to President Donald Trump’s speech during a news conference at the G7 summit, Wednesday, June 17, 2026, in Evian-les-Bains, France. (Jeanne Accorsini/SIPA via AP Images)

Commerce Eased Its Block on Anthropic's Mythos, But Major Questions Remain

The Tech Litigation Roundup spotlights notable lawsuits and court decisions across a variety of tech-and-law issues.

This month, the US Supreme Court issued a decision in 

 that all but eliminated pathways for tech accountability under both the Alien Tort Statute (ATS) and the Torture Victim Protection Act (TVPA). Both laws concern liability for human rights violations, so the Supreme Court's decision restricts litigation over some of the most egregious allegations of abuse by US tech companies.

The lawsuit against Cisco claimed that the tech giant built bespoke products to help the Chinese government surveil and persecute a religious minority, the Falun Gong. According to the plaintiffs, Cisco designed an internet surveillance system called the “Golden Shield” and a facial recognition-equipped video surveillance system that enabled China to identify Falun Gong practitioners, arrest them, and subject them to torture during forced-conversion sessions. In her 

, Justice Sonia Sotomayor highlighted media reports that thousands of Falun Gong believers have been tortured to death.

As Justice Sotomayor recounts, plaintiffs allege that Cisco explicitly marketed its services as “useful to the ‘

’ [crackdown] of Falun Gong” and characterized Falun Gong and its members as “viruses” and an “evil cult” in internal materials. In addition to helping China identify and track Falun Gong practitioners, Cisco’s surveillance systems allegedly provided information that Chinese officials used to threaten torture victims’ family members to coerce them into renouncing their beliefs.

Falun Gong adherents sued Cisco and two of its executives based on the ATS, alleging that they aided and abetted violations of international law by assisting in this campaign of persecution. One US citizen plaintiff also alleged that the executives aided and abetted his torture in violation of the TVPA.

US tech liability under the Alien Tort Statute pre-

, enacted by the First Congress in 1789, gives federal courts jurisdiction to hear cases by “alien[s]” (foreign plaintiffs) for torts “committed in violation of the law of nations or a treaty of the United States.” For nearly two centuries, it was 

, the ATS saw a resurgence as a basis for victims of serious human rights violations to sue the perpetrators in US courts. In 2004, the Supreme Court endorsed the ATS's use for this purpose in 

 while cautioning that “independent judicial recognition of actionable international norms” should remain “subject to vigilant doorkeeping.”

Since that relative high point for the ATS, however, the Supreme Court has repeatedly limited the law’s scope. In 

, the Court held that the ATS did not apply to conduct that occurred entirely in other nations’ territories. In 

, the Court held that foreign corporations could not be liable under the ATS. In 

, the Court held that allegations of “general corporate activity” occurring within the US were insufficient for the ATS to apply. All of these restrictions led some to 

 itself demonstrates, though, US Big Tech remained unusually vulnerable to liability under the already-weakened ATS. As American companies, these tech giants were not immunized under 

. And the nature of digital technologies means that employees in the US can sit at desks in Silicon Valley providing technical support for persecution across the world, potentially exposing them to liability despite the ATS’s exclusively domestic application. 

 that the ATS––even with significant carveouts––applied to Cisco’s alleged conduct, suggesting that the ATS could still be a meaningful tool for accountability when US tech companies contract with governments committing human rights abuses.

decision, that will not happen. Instead of continuing to chip away at the ATS––for example, by holding that it does not support aiding-and-abetting liability––the Court issued a broad ruling that renders the statute nearly useless.

To understand the decision, it is important to know that the ATS is a jurisdictional statute that does not itself create any causes of action. That means that the law gives federal courts power to hear disputes, but it does not create any new power for plaintiffs to sue. Plaintiffs either have to rely on some other congressionally enacted law that authorizes them to sue, or they have to ask the court to recognize a judicially-created cause of action.

That unusual structure, the Supreme Court recognized, is the result of a completely different conception of the law at the time that ATS was passed. In the 18th century, certain causes of action were understood to exist within “general common law” for federal judges to “discover.” General common law has since been eliminated from US law altogether, and causes of action must now be created by Congress or judges. Importantly, this Supreme Court has been hostile toward the idea that judges should be the ones to create causes of action. The Court already displayed its skepticism when it 

, a key judicially-created cause of action used to challenge violations of US constitutional rights.

, the Supreme Court held that federal courts lack the power to recognize causes of action under the ATS. To be more exact, courts theoretically have this power, but any conceivable cause of action that could be recognized under the ATS would violate the stringent test already used to decimate 

As a result, the range of viable ATS causes of action is––in Justice Amy Coney Barrett's words––a “null set.” This technical distinction allowed the Supreme Court to abandon 

's holding without explicitly acknowledging that it was overruling the case, prompting Justice Sotomayor to call 

 “yet another low point in this Court’s esteem for its precedents.”

In practice, this means that there are only two types of cases that can be brought under the ATS. The first is cases where Congress has passed a separate law authorizing suit, as with the TVPA. Because federal courts already have jurisdiction over cases 

, the ATS will generally be superfluous in this context. But even assuming the ATS has value here, this pathway is unlikely to be useful for tech accountability litigators. As discussed below, 

's narrowing of TVPA liability likely prevents tech executives' exposure, and Congress does not appear poised to pass new legislation that would welcome victims of human rights violations into US courts to challenge tech corporations' complicity in their abuse.

The second type of case that can still be brought under the ATS are challenges based on the “Blackstone three,” the set of torts that the Supreme Court has concluded were actionable under the ATS at the time of its passage: violations of safe conducts, infringements of the rights of ambassadors, and piracy. These archaic torts similarly offer little to those in search of tech accountability. Ultimately, the holding effectively eliminates the ATS as a vehicle for human rights and tech litigation, as well as for 

contemporary human rights litigation more broadly

TVPA holding likely precludes tech executive liability

In addition to the ATS claims, one plaintiff in 

 alleged that two of the company’s executives violated the 

The TVPA imposes civil liability on certain people who “subjec[t] an individual to torture.” The statute only provides for 

, so corporations cannot be sued under the TVPA. But in cases like 

, where plaintiffs have enough information to allege facts about how individual executives participated in deals that led to their torture, the TVPA still appeared able to meaningfully further accountability for alleged Big Tech participation in gross human rights violations. Given executives’ individual exposure, the prospect of TVPA liability may have been particularly effective at shaping the decision-making of the people who sign off on contracts that facilitate state violence.

 put an end to that possibility. The Court held that there is no aiding and abetting liability under the TVPA, so executives who knowingly provide technology that help state actors commit torture are not liable under the statute. This conclusion is narrower than the Court's ATS holding, and the TVPA will still provide a meaningful remedy to torture victims in many contexts. But against tech executives, whose participation in acts of torture will virtually always be at least as indirect as the alleged acts of Cisco executives, the holding is likely fatal to the TVPA's prospects as a tool for accountability.

US tech and human rights litigation after 

’s restrictions on the ATS and TVPA arrive at a moment where US tech companies are becoming 

 with the US government, and TJL and others have raised concerns that the US partnership with tech is 

 and other grave harms. These legal and political developments have left the future of US-based tech and human rights litigation uncertain.

One possible outcome is a turn toward foreign courts for human rights litigation against US tech companies––though this strategy brings its own challenges. It may not always be possible to obtain jurisdiction over US companies in a given foreign forum, and evidence of US corporate complicity may be harder to access from abroad.

In the US, plaintiffs may turn to litigating foreign law claims in American courts or relying on state law torts to obtain remedies for human rights abuse. For example, in 2024, a jury in Florida 

found Chiquita Brands International responsible

 under Colombian law for financially supporting a paramilitary group that killed the plaintiffs’ relatives. And for litigators of ATS cases whose core claims failed after prior Supreme Court decisions narrowing the statute’s scope, common law tort claims originally included “

” have sometimes allowed the cases to remain viable. As tech companies face 

 may put the same sources of law to the test for human rights harms that would previously have been challenged under the ATS or TVPA.

's elimination of two key vehicles to challenge tech complicity in human rights violations leaves companies and executives less constrained by liability fears. Victims, meanwhile, are left searching for new pathways to redress.

SCOTUS says Constitution protects cell phone location data:

 In another major decision this month, the US Supreme Court held in 

that obtaining cell phone location data from tech companies counts as a search under the Fourth Amendment. As a result, police will generally need a warrant to legally acquire this data; the decision stopped short of determining that 

 are always constitutionally insufficient. The decision recognizes a reasonable expectation of privacy in information shared with a tech company under the Constitution, a significant step for the legal framework governing digital privacy.

OpenAI faces investigation and lawsuit by state attorneys general: 

The Florida Attorney General sued OpenAI on June 1 for alleged legal violations related to AI safety. The first-of-its-kind lawsuit raises some claims familiar from private lawsuits 

 and others––such as violations of an unfair trade practices statute, negligence, and strict product liability claims––and a novel claim of public nuisance, which is 

 in the social media context in New Mexico. Weeks later, OpenAI was 

 as part of an investigation by a coalition of state attorneys general.

After Mythos/Fable takedown, lawsuit challenges US power to restrict model access:

 A legal tech firm that relies on Anthropic’s LLMs for its own products 

, arguing that its order requiring Anthropic to restrict foreign nationals’ access to the company’s most advanced AI models was unlawful and exceeded 

 Proponents and opponents of the data center boom both took their fight to the courts this month. The US followed through on its threat to intervene in the 

 the Public Service Commission of Wisconsin to challenge financial security requirements regulating hyperscale data center operators in the state. Meanwhile, residents of 

 brought actions raising different legal theories that both challenge an alleged lack of public input in procedures to approve data center tax breaks in Hillsboro and construction in Box Elder County. In Mississippi, another lawsuit was filed challenging the gas turbines powering xAI’s data centers near Memphis. The 

 alleges private nuisance, public nuisance, negligence, and negligent infliction of emotional distress based on noise pollution from the turbines, taking a new approach from the 

environmental and procedural challenges currently predominating

International cases with major implications:

 India’s authority to temporarily but entirely restrict access to the messaging app Telegram against a constitutional challenge, 

 over impacts on freedom of expression. Separately, a German court 

 for false statements in AI summaries appearing at the top of search results, because––unlike a traditional search engine––Google, not any third party, is responsible for generating the text.

Litigation spotlights increasingly advanced scams:

 what it says is a network scam subscription app makers alleging they extensively use shell companies to dodge app store enforcement. Google 

 Doe plaintiffs alleging they run a cybercrime operation that leverages Gemini to create fake websites for phishing. And a proposed class action in California 

 that AI company Kalibrate helps gas stations illegally collude to raise prices for consumers.

Amazon sued over Ring’s Familiar Faces feature:

 over an opt-in Ring feature that allegedly subjects every person captured on the doorbell camera to facial recognition AI and stores their biometric data. While some states have statutes explicitly regulating biometric privacy, this lawsuit relies on more widely available causes of action including deceptive trade practices, privacy torts, and unjust enrichment, and seeks to represent a nationwide class.

Social media logos are displayed on a smartphone alongside an under-16 age-restriction symbol in this photo illustration, as governments around the world introduce or consider bans and stricter age-verification rules to limit minors' access to social media platforms. (Photo by Jonathan Raa/NurPhoto via AP)

 of social media access restrictions for teenagers, an eye-catching alternative to the Australia-style ‘hard’ ban has begun to gain traction within tech policy advocacy circles.

This idea of a ‘presumptive’ ban on social media — whereby platforms would be banned for teenagers 

 they can demonstrate to a regulator or similar such body that they are safe — satiates the political appetite for access restrictions while recasting them as a lever for safety-by-design.

As such, it is instinctively appealing for a policy community that has spent years advocating for platform accountability and which isn’t ready to throw in the towel.

Yet as with the policy approach that it seeks to displace, the idea of a presumptive ban has complications of its own. Most fundamentally, there are real questions as to whether the contemporary online safety regulatory apparatus is even 

 of administering this policy approach in the manner envisaged by its proponents.

alternative to a ‘hard’ ban, it’s time to put this proposal under the microscope.

In recent weeks, calls for a presumptive ban as an alternative to an Australia-style hard ban have grown louder and from more prominent places.

A coalition of over 140 civil society organizations 

 European Commission President Ursula von der Leyen, demanding ‘independent ex-ante risk assessment, age-tiered child safety pre-certification, and regular independent auditing for all services accessible to children’.

In the UK, the chair of the parliament’s digital committee 

 the country’s government to mandate that ‘any new [social media] service should demonstrate that it meets safety standards before going to market’. The Online Safety Act Network — a UK-based coalition of advocacy groups — 

 framework for how such safety standards could look.

And Canada is set to become the first country to put the presumptive ban approach on a regulatory footing, with its proposed Safe Social Media Act 

 by which social media companies can avoid a teen ban by demonstrating to the regulator that they are safe for children.

A proposal made for a regulatory world that doesn’t (yet) exist

Ironically, the greatest strength of this policy approach — its focus on pre-market scrutiny of social media’s risks — is simultaneously its greatest vulnerability.

By relying on a regulator (or any other independent third-party) to assess and certify that social media platforms are safe before they are made available to teenagers, the proposal makes an ask of the online safety regulatory apparatus that it is today ill-equipped to deliver.

Regulators’ structural aversion to risk

‘No one ever got fired for buying IBM.’

This celebrated adage, which explains how perverse organizational incentives motivate deficient decision-making in large corporations, is apt for the regulatory sphere.

The regulators responsible for administering the presumptive social media ban face a deeply skewed incentive structure that will have significant implications for how this policy would work in practice.

To see why, imagine for a moment that you are the official responsible for assessing whether a social media service should be certified as safe for use by teenagers, on the basis of a pre-defined standard for ‘safe’.

Certifying that the service is safe, when the default assumption is that it is not, is unlikely to bring any significant upshot for you or your regulatory organization — there are no brownie points for doing your job.

On the other hand, certifying a service as safe that is later found to pose unpalatable harms to children would be reputationally catastrophic for you as an official and for your organization as a publicly accountable body. It’s not the kind of achievement that will look good in an annual performance review.

Add into the mix the reality that you are unlikely to face any meaningful blowback from your superiors for 

certifying a service as safe (more than most organizations, regulators struggle to quantify the cost of 

taking a risk), and we begin to see why the assessment and certification process is likely to be less objective and less technocratic than assumed.

This concern is illustrative of a longstanding challenge within the discipline of risk regulation. Across other highly regulated domains — from 

 — there are countless examples of how a structural aversion to risk serves to undermine the work of regulators and the achievement of public policy objectives. This phenomenon has been 

in the context of banking regulation in the post-Global Financial Crisis era, and has become a central preoccupation for policymakers everywhere in the 2020s amid the elusive search for growth and productivity. Indeed, in a 

 on balancing risk and innovation in regulation, the UK government noted that its “[national] regulatory approach has become too risk averse. Incentives on politicians and regulators to avoid criticism when regulation fails to fully mitigate risks have encouraged excessive risk aversion in the system.

One might reasonably counter that today’s online safety regulators appear to be operating just fine, and that they are hardly laid low by chronic risk aversion — just look at all the DSA investigations that the European Commission has launched.

But that is because regulators’ responsibilities today are of a 

 character — namely, to identify and punish 

 with a codified set of (largely procedural) rules. This work clearly entails risk for the regulator, but by focusing on outlier behavior (the frameworks assume compliance until a regulator proves otherwise) and by anchoring that compliance in a set of codified rules and procedures, the risk to the regulator itself of getting things wrong is relatively low.

The presumptive ban approach, on the other hand, would shackle regulators with a 

 would be on the hook for assessing and determining whether social media platforms are safe for use by teens. No matter who codified the standards for safety, there will invariably be some degree of judgment call and assumptions on the part of the regulator in the assessment process. There’s a whole lot more jeopardy involved in that work, and as demonstrated above, the skewed incentive structure that shapes how regulators approach their work means they will face a strong bias against certification.

For those of us who would like the presumptive ban approach to provide a 

 pathway for safe social media services to be available for children, that’s a problem.

Key to the idea of a presumptive social media ban is that services are restricted until they 

that they are safe for children. For supervisors who have organized themselves for a regulatory model that focuses on non-compliance with formalized procedural rules, this would require a fundamental rethink of their technique and their tools.

When administering a presumptive ban regime, the overarching question for regulatory supervisors will no longer be ‘do we have evidence that this company is breaking the rules?’ It will instead be ‘are we sure this platform is a safe environment for teenagers?’

Answering the latter question requires a whole lot more data, understanding, and situational awareness of platforms than the former. ‘Safety’ is a holistic and outcome-focused concept, and it is unlikely to be meaningfully assured by

 whether platforms have completed a risk assessment in line with regulatory guidance or offered workable user-facing reporting tools.

Indeed, the shift in focus from non-compliance assessments to safety assessments will stretch regulators’ evaluation capacities and tools to their limits. To certify a social media service as safe, regulators will need to undertake significantly more probing of technical systems, assemble richer and more diverse trust & safety metrics, and develop a far deeper understanding of how risk manifests across different product surfaces and to different user groups.

The challenge of developing and maintaining evaluation tools and techniques is a feature of all regulatory domains that require pre-market scrutiny and authorization. But unlike sectors such as automobiles or foodstuffs, regulators in the social media space exist on the wrong side of a pronounced asymmetry of information, marked by a huge number of known unknowns and unknown unknowns. That asymmetry means it is already hard enough for regulators of social media to do their supervisory work today. But absent a suite of new and innovative tools and techniques, delivery of safety certifications under this novel regime won’t simply be hard — it will be impossible.

Is there any hope for the presumptive ban approach?

By now, a presumptive social media ban as a credible alternative to ‘hard’ access restrictions might sound fanciful. Yet while profound, the challenges facing it can be overcome.

To get there, four things need to happen:

First, we need a clear articulation of what ‘safe’ social media is and how it can be assured. Companies need to know what is required for them to demonstrate their service is safe, and regulators need a toolbox of precise metrics and performance standards to evaluate safety. 

There are numerous pathways to get there, encompassing everything from platform functionality (e.g., certain functions must be restricted for a service to be considered safe); safety processes (e.g., services must complete risk assessments that meet certain 

 standards of quality and rigor); or performance outcomes (e.g., services must fall below certain quantitative thresholds for harmful content, behavior, and experiences). Advocates have already begun to spell out how that might look and there is a rich body of work from across the tech accountability community on which to build.

Second, we need an ecosystem of safety assurers beyond regulators alone. The skewed incentive structure that will lead regulators to be overly risk-averse when conducting safety assessments is difficult to design out. But it 

 be mitigated by introducing polycentricity to the assessment process. 

These regimes should create space for independent third parties to perform assessments and review companies’ safety cases (on the basis of agreed standards), as happens in many other sectors wherein 

 safety tests are required. This will ensure regulators – through their propensity for risk aversion – don’t become the single-point-of-failure that undermines the viability of the model. The DSA’s conception of third-party auditing, alternative dispute resolution, and its provisions on data access for researchers have created the green shoots through which this kind of polycentric oversight model could grow.

Third, transparency. Under this policy approach, regulators will have tremendous power to enable or frustrate teenagers’ access to key online services. There will be a significant public interest in the decisions that regulators make about service safety (or lack thereof) and how they make those decisions. One of the major criticisms of online safety regulatory supervision in jurisdictions like the EU and the UK today is that it is 

That way of working will be untenable in a world where regulators are playing the role of safety certifiers and making value judgments of tremendous importance to teenagers and society. For this regime to work, regulators will need to be transparent about the safety cases companies furnish and about how they arrive at their decisions to certify or not. Meaningful transparency in this regard will help mitigate unnecessary risk aversion by regulators and the politicization of their work, and it will ensure that the value judgments and trade-offs that are inherent in these exercises can be scrutinized and challenged by experts, advocates, and those affected by them.

Fourth and finally, serious engagement with the broader policy impact. This piece has engaged with the feasibility of the presumptive social media ban policy approach, but has left the questions as to whether this approach is good in itself largely to one side. These substantive questions must sooner or later be tackled head-on. For instance, is it desirable in a liberal democracy for technocratic officials in agencies, rather than elected officials in parliaments, to be the ones deciding which social media platforms are appropriate for teenagers to use? And what about the risk that complex procedural certification processes will introduce new barriers to entry for upstart alternative social media companies, which lack the compliance, legal, and lobbying resources of today’s entrenched firms?

The Canadian Safe Social Media Bill — that will make its way through the parliamentary process in the coming months — provides a timely impetus to engage substantively with the challenges inherent in the presumptive ban approach, and to further develop mitigations for them.

Time is of the essence. The idea of a presumptive social media ban increasingly appears to be the 

 viable alternative to a ‘hard’ social media ban in many regions of the world.

By taking the step of turning this alternative approach from an abstract idea to a rigorous policy proposal, we have a realistic opportunity of ensuring teenagers’ continued use of social media that is safer and more age-appropriate.

When one of the United States’ leading artificial-intelligence companies told the federal government that its models could not be used safely for autonomous lethal weapons or the mass surveillance of US citizens, the government did not simply find another vendor. It branded the company a “

—and moved to cut off its ability to do business across the entire federal defense ecosystem. The President then 

ordered every federal agency to stop using its technology

That is not a procurement dispute. It is punishment for protected speech, dressed in the ordinary clothing of administrative discretion. In a forthcoming 

, now before a federal court in San Francisco, I call this phenomenon “

”: the executive’s use of routine administrative tools—funding, contracting, licensing, personnel decisions, regulatory designations—to penalize disfavored viewpoints while preserving the appearance of business as usual.

The danger of expressive governance is precisely that it looks lawful. Courts have long, and sensibly, deferred to the executive’s judgment about who receives a contract or a clearance. But that deference was built for genuine managerial choices, not for retaliation. When the government turns these low-visibility levers against a speaker, the constitutional violation can hide in the gap between what officials say and what they actually do.

And this company is not alone. The same template has been used against 

law firms stripped of access for representing disfavored clients

universities threatened with the loss of billions in funding

 over their programs and their faculty’s speech, against 

media organizations menaced with regulatory reprisal

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe