Frederike Kaltheuner

Frederike Kaltheuner is a German technology-policy researcher whose career traversed Privacy International's international investigations, the inaugural Technology and Human Rights Division at Human Rights Watch, the EU Commission's digital policy apparatus, philanthropic strategy at the European AI and Society Fund, and independent analysis. She is tracked here as a Voice because her public writing and editorial work constitute a distinctive European contribution to the movement's argument about AI: that the harm landscape operates on two planes simultaneously — through dangerous surveillance and power-concentrating AI systems, and through pseudoscientific AI systems that embed invisible normative assumptions while appearing objective. She is the editor and intellectual architect of Fake AI (Meatspace Press, 2021), the anthology that gave the movement its clearest framework for the second plane. See the linked Person entry for affiliations and career history.

Voice anchor

The slot Kaltheuner occupies in this corpus is the European civil-society-to-policy bridge voice: a researcher whose accumulated institutional authority spans civil society investigation, human-rights institution-building, EU policy advisory work, and philanthropic strategy — and who deploys that authority through public writing that challenges both AI-industry narratives and the European governance frameworks designed to address them.

Her distinctive contribution is carrying the civil-society movement's foundational AI critique — that AI harms through classification failure as much as through surveillance success, that pseudoscientific systems deserve prohibition not just mitigation — into the European industrial-policy debate, where those arguments are least welcome. The EU AI Act, the AI Continent ambitions, and the compute-buildout narrative that followed all presuppose that the problem with AI is the governance of a technology whose progress and value are taken as given. Kaltheuner's public voice consistently refuses that presupposition, and does so from inside knowledge of each layer — as investigator, rights-framework architect, policy adviser, and funder — that the argument must navigate.

Fake AI and the harm-from-failure argument

The corpus's primary movement treatment of AI hype as a rights concern is Fake AI (Meatspace Press, 2021), an anthology Kaltheuner edited and shaped intellectually. The book's central argument — that much commercial AI constitutes "snake oil," and that harm flows as readily from non-functional systems as from powerful ones — established a framework the movement's policy discourse has since absorbed.

The argument Kaltheuner elaborates in the Sciences Po interview has three components. First, automated classification systems create the appearance of objectivity while embedding normative choices — when those choices become invisible through automation, they become harder to challenge, and the misclassified individual bears the cost. Second, harm arises from failure as much as from power: the "Kafkaesque" situation where a malfunctioning AI system denies a person a benefit and the institution responds that the algorithm decided is not a less serious harm than surveillance — it is a different form of the same structural problem (unaccountable automated decision-making affecting real people's lives). Third, the appropriate regulatory response is not only risk mitigation on powerful systems but the establishment of "red lines" — domains where AI should not be attempted regardless of how well the technology works, because the underlying task (predicting gender from voice, inferring sexuality from facial structure) is pseudoscientific and harmful by design.

The third point — red lines as a regulatory concept — became an argument in the EU AI Act debate. The Act's prohibited-use list is a partial operationalization of this concept; Kaltheuner's framing of the intellectual case for it, through Fake AI and public talks at re:publica and elsewhere, contributed to the policy vocabulary the movement brought to that debate. Her earlier co-authorship of the 2018 joint report Privacy and Freedom of Expression In the Age of Artificial Intelligence with ARTICLE 19 at Privacy International established the foundational rights-framework scaffolding that Fake AI built on.

European AI industrial policy: the structural-dependency argument

Kaltheuner's current work, anchored by the EU AI Industrial Policy Monitor (with AI Now Institute) and an ongoing series with Tech Policy Press, constitutes the movement's most detailed European-specific challenge to the AI-sovereignty narrative.

Her core argument, most directly stated in "Is Europe Getting AI Wrong?", is that Europe is "structurally dependent on a very small number of US firms across the AI stack in ways that are really difficult to untangle." The dependency is not primarily a compute problem — it operates across the entire vertical stack (models, cloud services, application layers) in ways that make European companies' apparent vendor choices less competitive than they appear. The European policy response — massive compute investment, the AI Continent program — addresses the wrong layer of the problem.

Three elements of her framing are distinctive. Market structure before infrastructure: creating conditions for "truly competitive markets" — breaking vertical integration, ensuring interoperability, enabling vendor switching — must precede infrastructure investment, not follow it. Infrastructure investment inside an uncompetitive market structure deepens dependency rather than resolving it. Scenario pluralism: she critiques the single-trajectory assumption embedded in European AI policy — that frontier model advantages are decisive and that most economic value will concentrate at the frontier — while ignoring the realistic alternative where models commoditize or smaller systems prove economically sufficient. A high-stakes compute bet premised on one scenario is fragile industrial policy. Democratic legitimacy as a constraint: if European policymakers pursue AI industrial strategy by managing public opinion rather than genuine democratic deliberation, and the bet fails, democratic institutions will suffer the consequence — not just industrial competitiveness. This frames the governance failure as a democratic problem, not merely a policy-design problem.

The EU AI Industrial Policy Monitor, a Substack newsletter tracking European debates on competitiveness, sovereignty, and AI, with over 1,000 subscribers, extends this analysis into ongoing commentary. Her 2024 AI Now Institute piece "Public Interest AI for Europe?" applies the same structural-dependency lens to the question of what public-interest AI governance for Europe would actually require.

Institutional bridge function

Kaltheuner's movement significance partly derives from the accumulation of institutional roles that carry independent credibility in different registers. Her investigations at Privacy International established her as a practitioner of corporate data-exploitation research. Her founding of HRW's Technology and Human Rights Division established her as a rights-framework architect at one of the world's most credible human rights institutions. Her EU Commission advisory role placed her inside the policy apparatus she now critiques externally. Her founding of the European AI and Society Fund made her a primary source of philanthropic capital for European civil society doing AI-rights work. Each role gives her public voice a different kind of authority; the combination produces a researcher who makes the structural-dependency argument from inside knowledge of the industry, policy, civil society, and philanthropic layers simultaneously.

Expert witness testimony — European Parliament, UN, UK House of Lords, Belgian Federal Parliament — extends this credibility into formal legislative processes. Her analysis reaches German-language audiences through prime-time German television appearances and publications in POLITICO Europe and ZEIT Online; English-language audiences through Tech Policy Press, AI Now Institute, and the Bulletin of the Atomic Scientists; and specialist digital-rights communities through re:publica and AI for Good conference appearances. In 2026 she was named among Germany's most interesting women in AI.

Why this is a Voice entry

A Voice entry is created here, rather than additional structure on the Person entry, because Kaltheuner's public output is itself the load-bearing object the corpus needs to track: Fake AI as the primary anthology framing AI hype and pseudoscience as a rights concern; the EU AI Industrial Policy Monitor as the ongoing tracker of European AI industrial strategy from a structural-dependency perspective; and the body of public writing and testimony carrying the red-lines argument and the market-structure-before-infrastructure argument into European policy spaces. Without this entry, the corpus has no Voice anchoring the European civil-society-to-policy register — the specific position that challenges both AI-industry narratives and European governance frameworks from inside knowledge of both.