Newsletter: How to confront the threat of AI dictatorship

Good morning!

Happy Mother’s Day to all who play that role in someone’s life.

This week we have multiple pieces that focus on the challenge of confronting rising authoritarianism and the future of tech and democracy.

• Nathalie Maréchal, managing policy director at the Institute for Information, the Internet and Democracy at Northeastern University, writes that tech policy used to be a relatively bipartisan field. But positions that would be defensible in a different political context simply aren't at this time. Regardless of where you sit in industry, government, or civil society, the time to choose is now, she says.
• As I noted last week, RightsCon was effectively canceled last week after Zambia’s government demanded changes to the program amid pressure from China over the participation of delegates from Taiwan. It’s a stark reminder of the growing threat of transnational repression. For the Tech Policy Press podcast, I spoke to Access Now executive director Alejandro Mayoral Baños and the director of RightsCon, Nikki Gladstone, about their experience, why this moment matters, and what's next for the community they convene. Listen here.
• For another episode published this week, I spoke to scholar and former European Commission official Paul Nemitz, who is one of the authors of a new book titled The Open Future and its Enemies: How We Can Protect Free Society from AI Dictatorship. The book (available as a free PDF at the link) argues that three decades of under-regulation have produced the concentrations of wealth and power we now confront, and that the survival of democracy in the digital age will depend on citizens, civil society, and a new generation willing to treat their work as carrying responsibility not just for safety, but for fundamental rights and self-government.
• Tech Policy Press contributing editor Mark Scott writes that European policymakers are failing to grasp the seriousness of the threat facing the bloc's democratic institutions — and how to best meet these challenges under the bloc’s next EU budget, known as the Multiannual Financial Framework.
• Tech Policy Press fellow Petra Molnar spoke with a Kashmiri journalist who describes reporting under continuous monitoring, intimidation, and fear. Kashmir has become a dense site of surveillance, where biometric systems, CCTV, and other technologies and policies shape everyday life, she writes.

A number of themes discussed in the pieces above are expressed elsewhere on the site this week:

Digital regulation, AI governance & accountability

• Nicolas Spatola, researcher in social psychology at Artimon Perspectives and teacher at Sciences Po Reims and ESSEC, argues that "human in the loop” is becoming the default safeguard in public-sector AI, but presence isn't the same as judgment. Systems built for efficiency can push officials toward deference, reduce scrutiny, and make oversight a procedural formality.
• Emma Hatheway, a technology policy student at Columbia's School of International and Public Affairs, writes that the White House is reportedly weighing an executive order to vet frontier AI models before release. But review without consequences isn't real oversight — and processes co-designed with the labs under review aren't adequately independent, she writes.
• Kathrin Gardhouse and Amin Oueslati, senior associates at The Future Society, write that the EU AI Act, the most comprehensive AI regulation to date, was not written with agents in mind. “Our research suggests that the EU AI Act applies to agents in principle, but falls short in practice,” they write in a piece that analyzes five governance challenges AI agents pose and identifies key gaps for the EU to address.
• Laura Caroli, who works at the intersection of artificial intelligence and public policy, reports that after months of tense negotiations, EU legislators have reached a provisional agreement on the AI Omnibus. The deal delays enforcement of high-risk AI rules, adjusts obligations across sectors, and sets the stage for a broader battle over digital simplification, she writes.
• Thijmen van Gend, a researcher at the Dutch Organization for Applied Scientific Research and the Institute for Technology in the Public Interest, writes that even if the EU's age verification app and EUDI wallet improve data confidentiality and decentralize data to users' phones, their promise of empowerment falls short due to centralized critical functions. “In doing so, these trust technologies reshape a myriad of trust relations when people have to ‘prove’ their attributes online and offline,” he writes.
• Tech Policy Press fellow Liz Carolan reports that Ireland's regulator has opened a new Digital Services Act investigation into Meta over dark patterns. The probe comes amid rising scrutiny of Ireland's role as the EU's de facto tech watchdog and growing public demand for tougher platform regulation.
• In a second piece this week, Liz Carolan reports that Meta is harvesting employee activity to train AI agents that could replace them, raising questions about data exploitation, labor disruption, and whether privacy-focused EU rules can keep up.
• Faiza Saleem, a digital and technology policy consultant, writes that recent rulings on social media harms against children reinforce that risks are shaped by platform design rather than content alone, highlighting the need for GCC regulators to move beyond content moderation toward enforceable design-based accountability. “As global standards around platform accountability consolidate, the cost of delay is no longer just harm to users. It is also the loss of any meaningful say in how those standards are shaped,” she writes.

Censorship in India

• Tech Policy Press fellow and Internet Freedom Foundation founder Apar Gupta says India's online content blocking orders quadrupled in two years — from approximately 6,000 per year to 24,300 in 2025. In this piece, he examines the political economy of digital censorship and how AI panic became the latest accelerant. “Each expansion of state power is preceded, justified, or accompanied by a public anxiety,” he writes.
• Amber Sinha, contributing editor at Tech Policy Press, writes that recent developments in the Indian states of Karnataka and Telangana suggest a shift toward a localized patchwork of hate speech regulation that risks bypassing India's limited constitutional safeguards for free expression.

Competition, platforms & design

• Sumit Sharma, executive director of NextGen competition, writes that a US court found Google to be a search monopoly, but the remedies hinge on limited data sharing and syndication requirements — meaning potential rivals must still invest heavily to succeed. Real competition will depend on how these rules are implemented and enforced.
• The mobile platform debate often gets stuck on “open vs. closed.” The Foundation for American Innovation’s Joshua Levine proposes a sharper framework for thinking about mobile competition he calls the ‘Mobile Trilemma,’ in which privacy, security, and integration are competing axes — and gatekeepers can use any one as a shield or a weapon.
• Madeline Batt, legal fellow at the Tech Justice Law Project, offers the April roundup on tech litigation. This month covers a recent lawsuit (just voluntarily dismissed by the plaintiff) against Perplexity, Meta, and Google that alleges chatbot transcripts — including "Incognito" conversations — were shared for targeted advertising, turning intimate AI interactions into adtech data.

AI, geopolitics & the Mythos moment

• Mark MacCarthy, an adjunct professor at Georgetown University in the Graduate School’s Communication, Culture, & Technology Program and in the Philosophy Department, and Carl Schonander, a global technology policy professional and former diplomat with the US State Department, write that Claude Mythos confirms the need for cooperation on AI risk reduction. The Trump administration should use the occasion of next week’s Trump-Xi meeting to call for bilateral government dialogue on the issue, they write.
• Purvi Patel, an attorney and public health professional in the international humanitarian and development sector, and Nicole Bennett, a scholar and practitioner working at the intersection of migration, data governance, and digital technologies, write that the same tech companies increasingly build and sell the same core capabilities into humanitarian operations and into military and intelligence systems and then let the humanitarian use serve as a kind of reputational varnish. “Humanitarian agencies and their donors should require dual-use disclosure as a standard: if a company sells the same platform family for defense, intelligence, or domestic security, that must be disclosed up front, not uncovered through investigative journalism,” they write.

Labor, power, and the local impact of AI

• Adriana Abdenur, co-president of the Global Fund for a New Economy, writes that in Silicon Valley, AI is a race — who wins, who captures value, who deploys fastest. In Brazil, the question isn't who wins but how the race itself is structured. With elections looming and institutions strained, AI safety starts with everyday harms, not existential ones.
• Tech Policy Press fellow Lam Le reports that Singapore and Malaysia have rolled out platform worker laws to improve gig worker protections. But some workers interviewed by Tech Policy Press said “additional protections are meaningless if the algorithm is still designed to incentivize longer working hours in dangerous conditions,” she writes, raising questions about whether the reforms improve day-to-day conditions in the gig economy.
• Matt Williams, chairman of Sustain SJ, writes that no matter what tech CEOs say in shiny advertisements and in closed-door meetings, Big Tech will not solve the climate crisis. In Vineland, New Jersey, where Williams is a resident, community members are questioning a massive data center under development to serve Microsoft.
• Eli Lehrer, president of the R Street Institute, writes that creativity has always been bounded by tools, institutions, and coordination. If AI expands those constraints, it will not make humans obsolete. It will change what it means to create by making it far easier to turn ideas into reality.

What we're watching

In the US this week, Google DeepMind, Microsoft, and xAI agreed to provide the US government with pre-release access to advanced AI models for national security evaluations, as the White House considers a more formal federal review process for frontier AI systems. Separately, reports indicated that the US and China may have official discussions on AI risks and governance ahead of a potential Trump–Xi summit in Beijing. We welcome contributor perspectives on both developments, including the efficacy of the administration’s review approach and its implications, as well as the priority areas for potential US–China discussions on AI safety, governance, and deployment.

Something different

As I mentioned in this week’s podcast with Alejandro and Nikki, RightsCon's cancellation meant missing something we'd been looking forward to all year: time together with so many folks working on important issues from across the world. So we're trying something different. In the halls at RightsCon, we had planned to invite folks to help us put together a community cookbook of recipes from digital rights and democracy defenders across the community, gathered from the kitchens of the people doing this work. We couldn’t gather at RightsCon in person this year, but we can still set a table together.

Submissions are open—we hope you will send yours! And to get things started, here’s my great-grandmother’s recipe for a simple cake. If you told her you were coming to visit, she’d grab these store-brought ingredients out of the cupboard in her small apartment and whip up a cake. I’m thinking of her this Mother’s Day.

Send us your recipes!

-Justin