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?

June 6, 2026: Protesters and officers with the Department of Homeland Security clashed outside the Metropolitan Detention Center in Los Angeles on the first anniversary of increased US Immigration and Customs Enforcement (ICE) raids across Los Angeles County. (Photo by Ted Soqui/Sipa via AP Images)

As Google Scrubs ICE Detention Reviews, Platform Power Serves State Interests

Irish Minister for Foreign Affairs Helen McEntee delivers a speech at Dublin Castle in Dublin, during the opening ceremony of Ireland's EU presidency on Wednesday July 1, 2026. (Niall Carson IRISH Press Association via AP Images)

The Irish Presidency of the Council of the European Union Should Welcome Scrutiny from Academia

Illustration of a Tamagotchi with the OpenAI logo depicted on its screen. Tech Policy Press/Shutterstock

Illustration of a Tamagotchi with the OpenAI logo depicted on its screen. Tech Policy Press/

When Users Say 'Goodbye' to AI

A group of people walking around at the National Conference of State Legislatures. 

The 2026 National Conference of State Legislatures in Chicago. Photo by Justin Hendrix.

At NCSL, Geoffrey Hinton Warns State Lawmakers Not to Buy Big Tech's Anti-Regulation Pitch

UK Prime Minister Andy Burnham speaking during a visit to a family center in Sheffield, South Yorkshire, on Friday, July 31, 2026. (Annabel Lee-Ellis/PA Wire POLITICS)

Taking Stock of the UK Tech Policy 'Muddle'

Screenshot from a Google announcement depicting its launch of a Nano Banana 2 feature in Google Earth. 

Google Earth AI Fiasco Underscores Why Tech Firms Must Listen to Outside Experts

Peter Chapman is Associate Director with the Knight-Georgetown Institute (KGI).

Last month may come to be seen as a turning point for the EU’s Digital Services Act (DSA). Over the course of July, the European Commission paired major policy initiatives with a series of consequential enforcement actions involving some of the world's largest technology companies. Even seasoned 

 have been struck by the pace and scope of these developments.

Through enforcement decisions, the Commission began to clarify the DSA’s broad legal obligations. These actions will test Brussels’ ability to shape whether and how these companies redesign their products to better protect safety, information integrity, and user wellbeing amid a fraught transatlantic landscape.

July was a remarkably active month for the DSA. The Commission launched its annual report describing “

” risks identified by platforms and search engines operating in Europe and released the much anticipated report of its 

. These reports further clarified the Commission’s interpretations of risks and expectations for protection of minors under the DSA. But it was the Commission’s enforcement activity that marked the more consequential developments.

The Commission took steps to advance enforcement against multiple companies, with developments taking place throughout July:

 that Instagram and Facebook’s addictive designs breached the DSA;

 X’s action plan to bring its advertising repository and researcher data access mechanisms into compliance with the DSA, following a December 2025 

 AliExpress €550 million for failing to effectively assess and mitigate risks to European consumers from illegal products; and

 TikTok in breach of the DSA for failing to ensure safe accounts for minors.

Each of these decisions expect platforms to make design choices that address specific consumer harms – from exposure to scams and counterfeit or illegal products to addictive features that negatively impact the physical and mental wellbeing of users.

Financial penalties alone will not ensure safe designs. Over the last decade we’ve seen technology companies weather fines and penalties arising from 

 without fundamental change to their business practices. This makes sense given that Alphabet, Google’s parent company, brings in 

Preliminary findings and non-compliance decisions emerging from the Commission detail how European law requires platform design that protects consumers. In preliminary findings, the Commission 

 that TikTok “needs to change the basic design of its service” to prevent harm to the physical and mental wellbeing of its users, including minors and vulnerable adults. Meta “

 to implement design changes to both Instagram and Facebook” that effectively tackle physical and mental wellbeing harms to users.

But how? While the DSA provides the toolkit necessary for consumer protection, effective enforcement requires more clarity. Enforcement must ensure platforms deploy safe design and data practices, not simply penalize violations after the fact. It should demand transparency into important company choices, calibrated to specific audiences, including regulators verifying compliance, researchers assessing harm, or consumers choosing where to spend their time. And the Commission needs a clear, sustained approach to monitoring compliance and progress, not just at the moment of a decision.

These implementation challenges are not unique to European regulation. Regulators across a range of domains – including US technology enforcement agencies like Federal Trade Commission (FTC), the Department of Justice (DOJ), and state attorneys general – have long grappled with the same task now facing the European Commission: how to translate legal findings into lasting changes that meaningfully protect consumers.

” to address precisely that challenge. It serves as a forward-looking enforcement framework for US technology litigation. It was developed through a systematic review of nearly 100 prior remedies, including FTC and DOJ consent decrees, public health litigation, civil rights settlements, and technology-related cases, and offers lessons relevant to the Commission’s enforcement strategy.

In line with “Designing Technology Remedies,” effective enforcement should feature three interlocking strategies: harm prevention, harm mitigation, and governance. These strategies must complement each other. One strategy alone is unlikely to ensure the safe online experiences that consumers 

Harm prevention changes how companies design, develop, and deploy their products. This could include prohibitions on unsafe design features identified through enforcement, including 

, as well as safer default settings or limits on data collection, retention, and use.

Harm mitigation should empower users to effectively report, avoid, and respond to harmful experiences. 

 show that existing platform user safety tools have fundamental flaws. Enforcement implementation should require measurable targets for effectiveness – for example, related to ease of adoption, impacts on user behavior, and overall user satisfaction. Companies generate considerable data and maintain robust measurement infrastructure that should be used to demonstrate that tools work.

Governance addresses how companies make decisions and enforce compliance. The DSA establishes governance mechanisms that advance safety, including 

. These systems are not working anywhere near their potential. To effectively assess safety mitigations, platforms must do more than list policies. They must describe clear standards for how risks are assessed, including specific taxonomies of metrics related to specific risks. Independent researchers need real access to data, be it 

Finally, the Commission needs to establish rigorous protocols for credible independent monitoring of specific outcomes. More work needs to be done to prepare for monitoring Commission decisions. The Commission’s June 2025 acceptance of commitments from AliExpress, which establishes a 

 an emerging oversight model inspired by financial services. The Commission should ensure tailored monitoring and measurement of specific design-based enforcement, similar to what is 

 by US attorneys general in litigation against Meta. This should be done through a 

 of internal company monitoring and measurement as well as independent third-party assessment.

July did more than just accelerate the pace of DSA enforcement: it began to concretize how platforms can turn the DSA’s broad design principles into specific product features that advance safety. The Commission’s preliminary findings assert that platforms must alter core features like infinite scroll and personalized recommendation systems. Translating these findings into durable designs, backed by independent monitoring, remains the harder task ahead. Fines alone will fail to challenge business models built on mass data collection, economies of scale, and maximized engagement. For digital platforms to effectively address risks, the Commission’s next chapter must advance the onerous work of specifying, verifying, and enforcing how platform design can measurably reduce specific harm.

Last week, Brazil’s Electoral Court unveiled an 

 to force Big Tech firms to make their election efforts more transparent: compliance plans. In an unprecedented measure, platforms will have to provide comprehensive and structured information on how they monitor, address and counter disinformation, coordinated inauthentic behavior and AI-generated political content.

With just two months before Brazilians head to the polls, the question is whether the Electoral Court will have the means and willpower to enforce it—and if platforms will be willing to cooperate.

In October, Brazilians will head to the polls to vote again for president four years after the elections that put Luiz Inácio Lula da Silva back in the presidential office. Since then, his opponent, former president Jair Bolsonaro, 

was convicted and is currently serving a prison sentence

 for a coup d’état attempt which included a plan to assassinate political opponents. With Bolsonaro barred from running, his son, senator Flávio Bolsonaro, 

In the past two election cycles, Brazil’s Electoral Court has been seen as a stronghold against Big Tech and, particularly, the weaponization of social media by far-right groups. Since the 2020 elections, the court has 

 with rules pertaining to the use of social media during elections and has established cooperation agreements with academic researchers and platforms to ensure rapid crisis responses.

The Electoral Court’s protagonism in producing rules to regulate platforms around elections comes due to a vacuum of legislation, as Brazil has tried and failed to pass comprehensive social media bills in the last decade. In practice, this means 

measures restricted to the election periods very focused on combating information that could compromise electoral integrity, such as attacks on the electronic voting system.

But the new mechanism could shift the logic from this reactive stance to a more proactive posture focused on preventing and mitigating risks, said Tatiana Dourado, Professor at the Department of Communication at the Pontifical Catholic University of Rio de Janeiro and Director of Digital Policies at the Democracia em Xeque Institute, in what she sees as a normative improvement that has been evolving since 2019 leading up to 2026.

“There is a challenge, and a very positive one if it is met, of having platforms provide complete, detailed, comprehensive information,” said Dourado. “And there is another challenge on the TSE’s side on the capacity to follow and track each compliance plan provided by platforms,” she added.

In March, one of the resolutions was edited to include an obligation for platforms to present compliance plans, but the details of what these plans needed to include, what format and deadline were still pending. A 

in late July laid out the rules, giving platforms a deadline of August 16 to elaborate and present the plans.

The obligation of presenting the documents applies for platforms with over 5 million users in Brazil, either individually or together with other platforms from the same economic group, and excludes email providers and online meeting tools.

Among the requirements, platforms must detail their operational structures to receive, filter and meet court orders during the electoral period, the parameters and criteria used to identify “risk hypotheses” and coordinated inauthentic behavior.

It also demands that generative AI platforms describe the technical safeguards adopted to impede systems from recommending or favoring one candidate over another, to impede systems from generating sexual abuse images of candidates, and the mechanisms used to identify and label synthetic content in organic publications.

The decree determines that failure to meet the deadline could automatically subject platforms to a “regime of direct regulatory supervision.” Under this regime, the TSE would directly order requests for information, indicators and data. Another possible sanction against platforms who fail to deliver the plan would be the impossibility of running political advertisements, an 

“One of the differences [from 2024] is opening this window of opportunity for a regulation much more focused on risk prevention and mitigation, and one that demands more responsibility from digital platforms,” said Dourado, mentioning that this happens in the light of a 

 from the Brazilian Supreme Court on the intermediary responsibility regime.

The compliance plans will also offer academics, journalists and civil society organizations an unprecedented window into platforms’ methodologies and moderation efforts during elections. The decree divides the plans between a public layer and a separate, more detailed, layer limited to TSE servants and collaborators.

’ efforts to allow it to “become an object of ample scrutiny by society, especially those who follow Big Tech across time, especially during elections,” said Dourado.

This is already much more access than in previous elections, as platforms had been progressively shutting down windows of transparency. From public APIs to ad libraries, many had altogether discontinued measures that permitted academics, civil society and journalists to scrutinize their efforts.

“Transparency in 2024 was restricted to the need of a platform having an ad library. In 2026, the compliance plan demands platforms provide what is their risk assessment, it creates compliance measures for established risks and for the services offered with qualitative and quantitative indicators so they can be monitored,” the professor explained.

The Electoral Court is currently presided over by Justice Kassio Nunes Marques, appointed to the Supreme Court by Bolsonaro. That’s a stark contrast to the 2022 elections, when the court was presided over by Justice Alexandre Moraes, the magistrate largely seen as the responsible for Bolsonaro’s conviction. In addition to the cases against Bolsonaro, Moraes has also become known for his 

Since he took over the presidency of the TSE in May, observers have been following Nunes Marques’ measures. Some decisions have prompted more concern. In June, he granted a request from Flávio Bolsonaro’s campaign to 

that showed a drop in voting intentions for Bolsonaro’s son. The poll was the first released after the Intercept Brasil published a story about 

, a banker currently under investigation for what has been dubbed as the biggest financial fraud of the country’s history.

In May, Nunes Marques had edited a rule to allow himself and the vice-president of the Court, Justice André Mendonça, who was also appointed by Bolsonaro, to be rapporteurs of court actions about political advertisements. This change 

 of some of the most consequential actions that reached the court so far.

But the TSE isn’t the only different variable this time around. In 2022, when Brazil last held general elections, Big Tech were more cooperative and less aligned publicly with US President Donald Trump. They were also not at the center of an 

 waged by Trump against Brazil for supposed unfair commercial practices.

On the one hand, a softer stance from the TSE could have a positive effect on not escalating the geopolitical issue, said Andressa Michelotti, a PhD in Political Science at the Federal University of Minas Gerais, a researcher at Margem (Research Group on Democracy and Justice) and a member of Governing the Digital Society at Utrecht University.

“The question is what would things look like with a TSE like in the past, but with Trump now in power,” said Michelotti, referring to the court under previous presidencies that antagonized Big Tech more strongly.

But there is also a question, according to Michelotti, around the feasibility of implementing measures such as the ones required by the TSE’s new resolutions, particularly with smaller moderation teams and with the United States midterm elections taking place the month after Brazil’s vote.

“This is a hypothesis but we may end up seeing a scenario where removing is easier than keeping content online,” said Michelotti, speaking of the risk of excessive censorship by platforms to comply with deadlines and norms imposed by the court given the lack of staff resources.

Platforms’ willingness to comply with the rules is also an open question. While the compliance plans do not allow much leeway, there is a recent history of pushback. Earlier this year, when the Brazilian government 

 operationalizing the new Supreme Court understanding of the intermediary liability regime, Big Tech and the far-right 

But what is at play is much more than the risk of fines or other types of punishment. Brazil holds other assets that are very valuable for United States’ technology companies at this moment which could make them more willing to cooperate. “Brazil has leverage in this dispute, not only because of a significant user market—and we consider that users are data and this factors into their monetary and non-monetary calculations—but also because of data centers, this could become a negotiating token,” said the researcher.

 on the negotiation table between the country and United States trade authorities. A government-proposed bill with a set of tax breaks and incentives was approved in the Lower House of Congress in February and is now awaiting a vote in the Senate.

Twenty-nine countries, most of them from the Global South, signed the founding accord of the 

World Artificial Intelligence Cooperation Organization

 in Shanghai on July 16, convened by China to draw in governments outside the small group that builds frontier AI systems. On June 3, the European Commission adopted its 

 proposal, aiming to triple EU data center capacity within five to seven years and to reorganize the EU's internal market. Both treat AI infrastructure as a strategic question. Governments from the Gulf to Southeast Asia are doing the same, buying compute and funding national foundation models under the banner of AI sovereignty. The budgets are substantial. What they are supposed to purchase is often left unstated.

Sovereignty can be used to justify almost any purchase

Much of the current argument about AI sovereignty concerns whether it is desirable or attainable. Analysts debate definitions while ministers announce multi-year packages. Very little of this gives a finance ministry anything usable when two proposals compete for the same money. Sovereignty also comes in distinct kinds, including control over infrastructure, jurisdiction over data and the freedom to change suppliers. They often trade against one another. A budget office needs to know which kind it is buying.

The word sovereignty does little to guide that decision. Invoked loosely, it justifies almost any purchase. The argument also has an interested advocate. Since February 2024, Nvidia's chief executive Jensen Huang has told governments that every country must 

own the production of its own intelligence

. Nvidia sells most of the chips that such ownership needs. He may be right, but ministers should not take the vendor's word for it.

Full self-sufficiency across the AI stack — the set of infrastructure layers, platforms, tools and services that facilitate the use of AI — is out of reach, as the 

 argued in January. The Tech Policy Design Institute's June report, 

, counts 103 capabilities across six layers of the stack, from compute infrastructure to governance. No treasury can fund them all. Given a fixed sum, which capabilities should a government build at home, which should it pool with neighbors and which should it buy on terms that keep an exit open?

Governments should buy bargaining power rather than sovereignty

Henry Farrell and Abraham Newman supply the missing test. They 

 that bargaining power in the global economy comes from occupying positions in networks that other actors cannot leave at an acceptable cost. That is what leverage means here: what a government can credibly demand because walking away would cost the other side more. Such positions also weaken with use, because the squeezed side starts building routes around them, so the strongest leverage tends to be the kind rarely exercised. Taiwan's position in advanced chipmaking is one. Control over access to the European market is another. A capability justifies its budget line when it changes what a government can demand from suppliers, partner states or international bodies. A capability that many others also possess may serve other ends, but it adds no bargaining power, however sovereign it feels.

Before approving an AI line item, a finance ministry should ask three questions. Which stated national objective does the capability serve, and through what chain of steps? Growth, security and influence over global rules reward different investments. How much bargaining power does the capability generate? Can the country use it? A data center that no laboratory draws on produces little beyond electricity bills. And what would the same money produce in its next-best use, inside or outside the AI sector?

What did Switzerland and Japan actually buy?

 model is a subtler case. EPFL, ETH Zurich and the Swiss National Supercomputing Centre trained it on public machines and released it openly on September 2, 2025. Open release was the design goal, so Bern has nothing here to withhold or trade. That may be what its funders wanted. Publishing a model can also set standards that others adopt. That is a separate objective. It carries its own cost, which a budget should show. The model was trained on more than 10 million GPU hours of public supercomputing time. Run the test on that. Against commercial access to stronger models for a comparable outlay, Apertus buys less raw capability, more independence from closed providers and a model others can build on. A budget office should have to say which of those it is buying. Apertus belongs in the research budget or under standards-setting and not in the sovereignty column.

Japan shows the opposite case, along with its limits. Japan Investment Corporation, a state-backed fund overseen by the trade ministry, 

 in a $6 billion deal in 2024. Every leading-edge fabrication plant depends on photoresists, the light-sensitive coatings that carry circuit patterns onto silicon wafers. Few firms make them. Tokyo had already tested that position in 2019, when 

 on photoresists and two other chemicals pushed South Korea to find suppliers in Belgium, the United States and Taiwan. 

Set the two options against each other. The controls cost Tokyo no public money and produced the squeeze. They also taught the target to diversify. Ownership cost $6 billion and added little beyond what export laws already provided. The purchase secured equity in a position Japan already governed, regardless of the fund's stated rationale of industry consolidation. Then, in May 2026, Reuters reported that JIC was 

, with Fujifilm and Mitsubishi Chemical interested, after AI spending lifted valuations across the chip supply chain. Both suitors are Japanese, so the position would stay inside Tokyo's jurisdiction either way. Even so, a state that treats a scarce asset as something to be sold at the right moment has priced the position as an investment. Leverage that is for sale is already half spent.

Ministries that test for leverage find it in cheaper places

A finance ministry applying this test will also find cheaper sources of leverage. Power supply, water and permitted land increasingly decide where data centers can be built. A country that can offer all three has something scarce to trade, though scarcity alone does not settle who captures its value. A government that offers discounted electricity and a tax holiday hands that value to the investor. Host countries compete for the same projects, so many end up conceding it.

Generic datasets age the same way, as models get better at learning from less. Data keeps its value when it is current, legally usable and hard for others to assemble. Some locally held collections qualify: court archives, health records or recordings of a language few others have gathered. New Zealand's 

 shows how such data can be governed. The Māori organization built a speech dataset for its language and licensed it on terms that keep the benefits with the community. Governments can adopt the same principles of consent and benefit-sharing. Custody is not ownership, so they should not treat citizens' records as a sovereign asset to license at will.

Where one country's demand is too small to interest suppliers, pooled procurement or shared compute facilities can reach the scale at which suppliers negotiate. ASEAN, the African Union and the Gulf Cooperation Council could all do this, though today their members mostly negotiate with the same suppliers individually. The EU already pools for supercomputing through the 

Where a capability remains outsourced, the terms of purchase determine how much bargaining power a government retains. Multi-vendor contracts preserve an exit option at a fraction of the cost of building it at home, though only where migration has been tested. A buyer can also negotiate training quotas for its own engineers, the portability of models and data or the placement of source code in escrow with a third party. Governments often become dependent after the sale, once the only people who can run the system work for the supplier. The proposed Cloud and AI Development Act itself encourages public bodies to avoid relying on a single vendor.

Sovereignty is only one line in the cost-benefit table

Ministers will say, fairly, that capabilities get used in ways nobody forecast. A strict test would have rejected some investments that later paid off. Some spending is best understood as a financial option: a small payment now for the right to act later. But options have prices and expiry dates. Ministries can fund small, staged bets with capped downside, each with a date for renewal or closure. A government that cannot say what its options cost or what they might open up has not made an investment case.

Fund a capability at home when the leverage it generates toward a defined national objective exceeds the return on the next-best use of the same public money. Otherwise, pool it regionally, or buy it from the market while keeping an exit option. A sovereignty claim then becomes one line in a cost-benefit table, set against outcomes governments already pursue, such as jobs and a seat at the standard-setting table. The next sovereign AI budget that reaches a cabinet should arrive with an answer to one question: how much leverage does this money buy, and over whom?

appearing on Tech Policy Press in 2026. More from the series

AI images are a new form of structural violence—not through what we see, but through whom we 

 see. The gender violence I am referring to here is not about bias. It is about AI firms subtly but consequently rendering women invisible, while the biased stereotypes we do see online are just one of its symptoms. In this regard, the hype around AI operates as a gendered regime of visibility. As AI outputs circulate widely across society they normalize synthetically created images of women and their aesthetics—or make women disappear all together. Hence, AI hype not only frames public perception of emerging technologies in simplified terms but also simplifies the reality we see: erasing the grainy, the fluid, the non-conforming. In this way, it contributes to the erasure of women online.

 in ways that remain instructive today. Viewed through Arendt's distinction between power and violence, AI-enabled gendered violence affects the democratic life of women and the conditions under which they can participate within it. Arendt links power to visibility. Power emerges through people acting together in public and participating in shared political life. Violence, by contrast, diminishes that capacity. It interrupts collective action by intimidating, isolating, and silencing those it targets—making them invisible. Violence is not always spectacular. It often unfolds through ordinary gestures that become normalized through repetition.

As we debate and negotiate the dangers of online 

; an unfortunate and unwanted error the developers of these technologies apparently did not see coming. Similarly, it could be too simple to say that 

 itself causes gender-based violence. Not 

 AI application inherently contributes to gender violence, and not every user is a sexist antidemocrat. Rather, the AI hype and unbound enthusiasm surrounding generative AI have newly obscured questions of responsibility by presenting these systems as neutral tools whose effects depend entirely on the prompts provided by their users.

 in early 2025, many outlets and media were still predominantly focused on how deepfakes are jeopardizing 

 in early 2026 was a turning point in understanding gendered violence as a business model and part of the app design. Grok, a generative AI application developed by Elon Musk's company, xAI, was used to generate and circulate non-consensual sexualised images of women and children and share them on X and across other platforms, often with minimal intervention or filtering. These effects were widely framed as unfortunate mishaps: technical oversights, uneducated users, and failures of moderation—Musk even asserted the backlash was merely an “excuse for censorship” by regulators that sought to investigate.

 on Tech Policy Press, every decision about what a model and application like Grok does, permits, restricts, and prioritizes is co-produced by the companies and developers, not just the users nor the data. Systems enabling non-consensual image generation are not technological accidents. They were developed and deployed within environments in which such uses were entirely foreseeable. The question is therefore not simply why users exploited these systems. It is why infrastructures were designed in ways that tolerated predictable harms. Predictable or not, democratically speaking, Grok is a violent technology, not just a biased one. It might practically and virtually “undress” women, but on a wider scale, it inherently intimidates and silences all women due to its potential of violence.

In fact, technologies have never arrived free of values. Philosophers of technology have long argued that technical artifacts, not just AI, 

always embed assumptions, priorities, and socio-technical agency

. Similar, generative AI systems are not simply used for good or bad. They are shaped through countless design decisions involving training data, filtering mechanisms, shadow prompts, interface constraints, and optimization objectives. The conditions of their use are already structured before users enter a prompt. There is no neutral encounter with AI. AI is about power, and such is used to control the aesthetics of certain groups in society.

 of this series explains, hype is charged with values, worldviews and political positions. It is mobilized to attract attention, investment, and revenue, and the resulting perception of technology is therefore often simplified. Its very conditions of production—frequently shaped by consultants and marketing agencies—make it averse to nuance, unable to adequately address risks, and prone to rendering negative consequences invisible. Hype is not inherently problematic, but its objectives and underlying dynamics deserve careful scrutiny. We should remain attentive to who benefits from it, and who and what is rendered invisible, or harmed.

Synthetic discrimination: invisibility as structural violence

AI images are an interesting new form of media, but like any form of media they also 

, as Roland Meyer writes, which are opaque to users. Generative AI systems such as Grok, ChatGPT, and Midjourney are built upon complex entanglements of data, computation, and optimization. Data matters, but it is not only used well or not found. It is curated, selected, filtered, transformed, and reinforced 

through institutional and corporate choices

. AI images are derived from calculations and not necessarily documentation of real events, like historical images showing scenes or people that existed. This is not to say public or historical images have ever been neutral or immune to manipulation, but something has changed with AI. An AI-generated image of a woman can be created without any one real woman as its visual or representational ground.

Ultimately, AI favors its makers' instructions and not the voice or empowerment of its ‘objects’. AI images are not simply 

 images drawn from representations of existing women; rather, they are renderings of woman-like features assembled from millions of data points. Worse, they replace existing women’s images. Even if an AI generated image is drawn by reference to a real woman, once it is edited through prompts, datasets, statistical associations and platform values, it is detached from any particular subject.

This process is visually and democratically consequential. It leads to what I call 

. This dynamic builds on the idea that even if AI only creates images of women synthetically, these replace existing women’s images online (at first). Besides AI images relying on mostly male designs and developers, 

AI images are already flooding the internet

. One might object here and say that this dynamic should affect men, too, and maybe negatively. This is true, but not to the same extent. Also, men are not systemically devalued, violated or sexualized online and offline; and they 

are dominantly represented in AI design and data

 is an ongoing struggle, but it is getting harder. Older women are already less represented in search results and recommendation systems; women with disabilities may continue to be more absent from dominant visual cultures; and women of color may appear even more through stereotyped templates, though not as frequently as white women. These dynamics will now worsen as synthetic imagery multiplies. And now, we have to fight over seeing 

. Once online media formats or corporate outlets use images from image generators and not from photographs, women will become even more invisible due to the synthetic images replacing them online.

The democratic concern: pushing women out of (online) publics

Synthetic discrimination refers to more than only a failure to represent women fairly—it hinders the very condition that makes democratic representation possible. Core ideas behind democracy, including participation and visibility, are rendered impossible.

As a form of structural invisibility, AI is violent, not just biased. Violence, here, is then more about rendering women structurally invisible, and less about sexualizing them. Not to say that the second is not hugely playing into the first. This is a paradox we face with AI: It can violate women’s visual representation by not even

representing them as a subject while keeping them inherently vulnerable.

Referring back to Arendt, democracy lives from citizens contributing to any form of deliberation and to public debates through being recognized as legitimate speakers. Synthetic discrimination could hinder this process. What seems like an online issue will become one offline, too. The synthetic AI woman is not a political agent representing one specific, or 

, woman, but representing an ideological and aesthetically charged object produced by the network and powers behind AI.

If we permit digital publics to become saturated with hyperreal AI women—considering the internet is already 

—the representative space for women will keep shrinking. For now, synthetic discrimination is not the dominant online reality, but we shall not overlook the historically tight and concerning link between 

AI technologies are fundamentally anti-democratic

Synthetic discrimination manifests an urgent warning when it comes to violence, one that Arendt could have raised too: it is an accumulation of subtle, seemingly accidental and gradual gestures masked as technical or procedural errors. But again, the technical or visual level is not least political. Applications like Grok function as disciplinary mechanisms, as reminders of who controls participation, whose bodies remain vulnerable to public violation, and whose presence in digital spaces is conditional. In this sense, we cannot refer to Grok as only 

, because this is about body politics. Whether articulated explicitly or embedded structurally within design practices, surely the 

 club in Silicon Valley tends to see women less as constituencies or citizens to be represented but more as obstacles to be managed. 

AI-enabled gendered violence should not be understood as a narrow technical controversy or matter of biased data. This conceals the global and monopolistic scale which powers AI. The more these infrastructures become embedded within everyday life, the more urgently democratic societies must ask what forms of participation they make possible, whose visibility they secure, and whose disappearance they render tolerable. Protecting women from AI-enabled violence and invisibility is therefore not only a matter of safeguarding one vulnerable group. It is a democratic imperative. Addressing AI hype in order to protect democratic institutions and values requires acknowledging its regimes of visibility and invisibility equally, and not only those of bias. This means recognizing that, while hype flattens public perceptions of AI, it also enables new, subtle, but consistent, ways in which women are not only turned into sexualized as objects, but potentially into invisible citizens.

Thanks to some amazing women who motivated this research, to Eryk Salvaggio and Christo Buschek for inspiring thoughts, and to Andreu Belsunces and Jascha Bareis for their great feedback.

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