Manipulation without Lies: AI Hyperpersuasion and the Limits of the Deception Paradigm
Abstract Large language models can now outperform human persuaders in controlled settings, and they do so in whichever direction they are instructed to persuade. Regulatory and ethical responses to this development nevertheless remain anchored in what this article calls the deception paradigm: the assumption that communicative influence becomes wrongful when, and because, it involves false assertion. The article argues that this paradigm systematically fails to capture AI hyperpersuasion. Drawing on the philosophical literature on manipulation, it shows that hyperpersuasion characteristically operates through true but selectively curated content, adaptive framing, simulated intimacy, and the exploitation of individually diagnosed cognitive dispositions – techniques that bypass rather than deceive rational agency. It then examines Article 5(1)(a)–(b) of the EU Artificial Intelligence Act. Although the prohibition is conceptually open to manipulation without deception, its operative test – material distortion, appreciable impairment of informed decision-making, counterfactual decision change, and significant harm – is inherited from unfair commercial practices law and reproduces the architecture of fraud, leaving paradigmatic cases of hyperpersuasion untouched. Extending the analysis from individual autonomy to democratic epistemology, the article contends that hyperpersuasion threatens democracy not primarily by aggregating individual harms but by degrading the epistemic conditions – independence of judgment formation, a shared space of reasons, and the contestability of public claims – on which collective self-government depends. Regulating hyperpersuasion therefore requires a shift from content-based tests of falsity and harm to process- and relationship-based criteria that target asymmetric epistemic power.
Authors
- Fabian Maximilian Johannes Teichmann (ORCID: https://orcid.org/0000-0001-6543-6401)
Institutions
- London School of Economics and Political Science (GB)
Publication Details
- Journal
- Philosophy & Technology
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1007/s13347-026-01245-8
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00