Who is liable when AI meets human brain? Examining BCI responsibility gap

Brain–computer interfaces (BCIs), which translate neural signals into digital commands that control external devices, are becoming increasingly sophisticated as machine learning systems improve their ability to interpret brain activity. However, the growing influence of AI in these systems is raising complex legal questions about who should be held accountable when technology influences human cognition or causes harm.

The issue is explored in the study “WHO is responsible? Towards the normativity of AI-driven BCI technologies in product liability in healthcare,” published in AI & Society. In their research, the researchers analyze how emerging AI-driven neurotechnologies challenge existing product liability frameworks and introduce a conceptual model designed to explain how legal responsibility evolves in response to advanced technological systems.

BCIs and the rise of algorithmic influence

BCIs represent one of the most ambitious developments in modern healthcare technology. These systems collect neural signals from the brain and translate them into digital commands capable of controlling external devices. Patients with paralysis can use BCIs to move robotic limbs, type messages, or interact with digital systems without traditional physical input. In rehabilitation contexts, the technology can also support neuroplastic recovery by enabling patients to retrain damaged neural pathways.

AI is vital to the operation of these systems. Machine learning algorithms interpret complex neural data patterns, converting brain activity into commands that control devices. As these algorithms improve, BCIs are becoming more responsive, accurate, and adaptable to individual patients.

However, the integration of AI introduces a new dimension to medical device behavior. Algorithms do not simply execute fixed instructions. Instead, they adapt to incoming data, adjust their outputs, and refine their predictive capabilities over time. This capacity for continuous learning means that AI-driven BCIs can evolve after they are deployed, potentially altering how they interact with users.

The study highlights that such systems create what researchers describe as algorithmic normativity, a concept that refers to the way algorithms embedded within technological systems can shape behavior and influence decision-making processes. In the case of BCIs, algorithmic systems may influence how neural signals are interpreted, how commands are generated, and how users interact with the device over time.

This raises an important question: when AI algorithms influence neurological behavior or decision-making patterns, should the resulting outcomes be considered purely technical malfunctions, medical risks, or forms of algorithmic governance that carry broader legal implications?

The study argues that traditional product liability frameworks were never designed to address technologies that operate at the intersection of artificial intelligence and human cognition. Conventional medical devices function according to stable engineering designs, whereas AI-driven BCIs involve ongoing feedback loops between human neural activity and adaptive algorithms.

Because of this complexity, determining responsibility for harm caused by these systems becomes significantly more difficult.

Product liability law struggles to keep pace with AI neurotechnology

The researchers examine how modern legal systems are attempting to respond to the growing complexity of AI-driven healthcare technologies. In particular, the study analyzes the European Union’s updated Product Liability Directive, a regulatory reform designed to modernize liability rules for digital and AI-based products.

The directive represents a major shift in how liability is applied to advanced technologies. Under traditional frameworks, injured individuals often had to prove that a manufacturer acted negligently in designing or producing a defective product. This requirement created significant obstacles in cases involving complex digital systems, where identifying the precise cause of failure can be extremely difficult.

The updated directive introduces several important changes that could reshape how liability is assigned in cases involving AI-powered medical technologies.

One key reform is the explicit recognition of software and artificial intelligence systems as products under liability law. This means that AI-driven technologies, including algorithms embedded within BCIs, can fall under strict product liability rules. In such cases, individuals harmed by defective systems may not need to prove negligence; instead, they only need to demonstrate that the product was defective and caused harm.

Another significant reform involves the expansion of responsibility across the product lifecycle. In complex AI ecosystems, multiple actors may contribute to the development and operation of a technology. Hardware manufacturers design the physical interface, software developers build algorithms, data scientists train models, and service providers may modify systems through updates or maintenance.

The directive acknowledges this distributed innovation model by extending potential liability beyond the original manufacturer to include other actors involved in the creation or modification of digital products.

The study also notes that the revised legal framework attempts to address the problem of algorithmic opacity. Many AI systems operate as “black boxes,” meaning their internal decision-making processes are difficult to interpret even by experts. This opacity can make it extremely difficult for injured parties to identify the technical cause of harm.

To address this challenge, the directive introduces mechanisms that allow courts to rely on presumptions of causation in situations where it is excessively difficult for claimants to prove how a defect occurred due to the complexity of AI systems.

The new legal framework expands the definition of compensable harm to include certain medically recognized psychological injuries when linked to physical harm. This development is particularly relevant for neurotechnology, where systems interact directly with brain function and may influence cognitive or emotional processes.

Reflexive normative cascade and the future of AI responsibility

The study introduces a theoretical framework designed to explain how responsibility evolves in response to emerging technologies. The researchers call this framework the Reflexive Normative Cascade.

The concept describes a multi-stage process through which societal expectations, technological experiences, and legal frameworks gradually interact to shape accountability.

In the early stages of technological development, ethical concerns often emerge before formal legal rules exist. Researchers, policymakers, and the public may begin raising questions about privacy, autonomy, mental integrity, and the potential misuse of emerging technologies.

As technologies move into real-world use, practical experiences begin to accumulate. Patients, clinicians, and technology developers encounter unexpected risks, design challenges, and ethical dilemmas. These experiences generate societal pressure for clearer governance frameworks.

Eventually, legal systems respond by codifying new rules that attempt to regulate the technology and assign responsibility when harm occurs. In this way, ethical concerns gradually evolve into formal legal standards through a cascade of normative developments.

The study argues that AI-driven brain–computer interfaces are currently moving through this process. While legal systems are beginning to recognize digital products and AI algorithms within liability frameworks, the complexity of neurotechnology continues to challenge existing regulatory approaches.

One of the most difficult issues involves continuous software evolution. AI-driven systems can be updated, retrained, or recalibrated after they are released. These updates may significantly alter how the technology behaves, raising questions about whether liability should remain with the original manufacturer or shift to the entity responsible for the modification.

Another challenge involves the long-term influence of algorithms on human cognition and behavior. Because BCIs interact directly with neural signals, they may shape patterns of neural activity or behavioral responses over time. Determining whether such influences should be considered product defects, therapeutic outcomes, or user adaptations represents a complex legal and ethical question.

The researchers also highlight the importance of transparency and traceability in AI development. Ensuring that systems maintain detailed documentation about their design, training data, updates, and operational decisions may become essential for assigning responsibility when problems occur.