Contemporary AI governance typically treats ethics and law as complementary normative domains with different levels of enforcement. This paper argues that this distinction alone is insufficient given the evolving role of AI systems as artificial moral agents. Conceptualizing ethics and law as functionally distinct yet interdependent domains, the authors argue that the notion of functional agency to describe how AI systems generate normatively relevant outcomes through decision substitution, the embedding of societal norms in the design of AI systems, and behavioural steering. Drawing on case studies of recommender systems, large language models, autonomous driving, and care robotics, the paper demonstrates the systematic displacement of micro-level ethical reasoning by standardized, societal logics that tend to prioritize macro-level considerations such as beneficence over individual autonomy. This shift poses a structural challenge to human rights, which explicitly protect individual moral deliberation. The paper therefore calls for a paradigm shift in AI governance from embedding societal norms in AI systems to governing the construction of normativity itself grounded with an emphasis on the protection of individual ethical reasoning and stronger acknowledgement of moral pluralism.

Acceleration ethics addresses the tension between innovation and safety in artificial intelligence. The acceleration argument is that risks raised by innovation should be answered with still more innovating. This paper summarizes the theoretical position, and then shows how acceleration ethics works in a real case. To begin, the paper summarizes acceleration ethics as composed of five elements: innovation solves innovation problems, innovation is intrinsically valuable, the unknown is encouraging, governance is decentralized, and ethics is embedded. Subsequently, the paper illustrates the acceleration framework with a use-case, a generative artificial intelligence language tool developed by the Canadian telecommunications company Telus. While the purity of theoretical positions is blurred by real-world circumstances, the Telus experience documents acceleration AI ethics as a way of maximizing social responsibility through innovation.

This article warns against the idea that information from neural speech decoding (NSD) can be treated as testimonial evidence. We contend that NSD and testimony differ in their epistemic status. While the evidential validity of testimony depends essentially on someone’s sincerity, by contrast NSD depends upon the accuracy of decoding the person’s inner speech. To show this, we first explore how NSD poses a problem for rights against self-incrimination by testimony in, for instance, the Fifth Amendment to the US Constitution. Building on Pardo’s epistemic account of the distinction between physical and testimonial evidence, we argue that since NSD produces ambiguous information, it does not qualify as testimony. Conversely, denying our claim makes one liable to what we call the informational symmetry fallacy (ISF), viz. the conflation of decoded output from someone’s brain with that person’s speech act. We then set out the consequence of falling into the ISF by criticizing Farahany’s proprietary conception of thought. Her conception is shaped by goal of providing a new taxonomy of evidence better suited for relating constitutional rights to brain data and privacy. Here, we argue that Farahany must presume that we already own and are thereby accountable for all of our intercepted thoughts. Hence, she precludes the possibility of our disclaiming some thoughts and thereby justifies others in treating those thoughts as our beliefs, without considering the dependence of belief our epistemic authority. Lastly, we discuss a practical problem entailed by our account, namely that users of neural speech prostheses will not qualify as witnesses in a legal setting. In response, we argue that although NSD prosthetics enable users to communicate, there are insuperable limits in assessing someone’s testimonial sincerity solely on the basis of an NSD readout. For this reason, we conclude that NSD is better analogized to hearsay evidence.