Aug 2026· Synthese· Vol 208· 1 citation· 36 references
Abstract
Our reliance on the new wave of generative AI has increased exponentially in the last few years, with little sign of abating. This raises the natural question of whether we should trust this technology as a source of information. Drawing on the epistemology of trust, it is argued that generative AI does not meet the conditions for being a trustworthy source of information. In particular, it is argued that generative AI is untrustworthy because it is an unsafe source of information. This, in turn, entails that trusting generative AI as a source of information is not a route to knowledge.
Is the human predictable by AI, and if so, should we allow it? This paper examines the relationship between these empirical and normative questions from the perspective of the right to free development of personality enshrined in Art. 2 of the German Constitution. Based on an analysis of 50 decisions of the German Constitutional Court, I advocate for the recognition of a right to unpredictability derived from Art. 2(1) of the Constitution. I argue that, even if humans statistically follow patterns, the freedom to act against all probabilities is worth protecting.
It is argued that both AI and bullshitters are untrustworthy informants, and for similar reasons, it is natural to describe AI’s informational outputs as bullshit, as it signals their distinctive kind of epistemic deficiencies, which they share with bullshit.
The Generative AI revolution is driven by corporations demanding legal superpowers. If we allow it to continue unchecked, the implications will be profound. This urgent, critical book exposes the unprecedented push by trillion-dollar companies to build AI on billions of unauthorized human works and redefine fundamental areas of law, including copyright, contract, and free speech. Written by an industry insider who turned from AI champion to AI critic, this highly accessible work promotes AI literacy and provides essential tools to pierce the hype. Readers will learn how to assess AI's profound societal risks to democracy and autonomy and ensure that we are the architects of-and not bystanders in-our artificial future.
Generative AI does not merely produce biased outputs. It encodes the majority's way of knowing as the default infrastructure of knowledge itself. We call this epistemic subordination. The training process compresses the full breadth of human expression into a single probabilistic model whose statistical baseline reflects the languages, assumptions, and cultural frameworks of the dominant culture. Minority epistemologies are not excluded but absorbed: present in the training data, yet structurally subordinated in the output. The result is not a collection of discrete biases that can be audited and corrected. It is an epistemic condition embedded in the architecture from which all outputs emerge. This unified harm cuts across three legal domains -- anti-discrimination law, cultural and linguistic rights, and democratic viewpoint pluralism -- and each fails to address it for the same structural reason: existing law regulates downstream, at the level of decisions and applications. The remedy must match the site of harm. If epistemic subordination is produced at the level of model training, then law must learn to govern at that level.
Practices such as “p-hacking” or “fishing” are widely regarded as epistemically pernicious, raising a puzzle. If two researchers generate identical data sets, why should it matter that one “fished” for the result whereas the other predicted it? Some respond to this puzzle by abandoning the Principle of Total Evidence (PTE), proposing that we exclude “tainted” data or weaken our descriptions of evidence. I argue that such departures are unnecessary and counterproductive. The most common forms of p-hacking are pernicious precisely because they violate PTE. For the remaining cases, I demonstrate that suppressing evidence leads to absurd results.