Human rights law provides a critical foundation for AI governance, offering established standards across substantive principles, enforcement processes, and accountability mechanisms. While AI technology has evolved at a rapid pace, governance frameworks remain underdeveloped, leaving significant gaps that human rights norms can help address.
In recent years, ethical principles for AI have proliferated across corporate, civil society, and intergovernmental domains. Common themes—data protection, transparency, bias mitigation—are widely cited, yet their definitions vary, and abstract notions like beneficence or non-maleficence often lack practical applicability. Philip Alston captured the challenge succinctly: “as long as you are focused on ethics, it’s mine against yours. I will define fairness, what is transparency, what is accountability. There are no universal standards.”
Privacy emerges as one of the most pressing concerns. AI’s appetite for vast datasets drives unprecedented collection and sharing of personal information, often without adequate consent. High-profile breaches include the unauthorized transfer of 1.6 million UK patient records to Google’s DeepMind, Cambridge Analytica’s harvesting of up to 87 million Facebook profiles, and Clearview AI’s facial recognition database built from 10 billion online images, found unlawful in multiple jurisdictions. AI’s profiling capabilities can infer sensitive attributes—sexual orientation, health status—without individuals’ knowledge, enabling micro-targeted advertising, political messaging, or even identity theft. International human rights law requires data processing to be lawful, transparent, and secure, with heightened safeguards for sensitive data. The EU’s GDPR exemplifies privacy protections rooted in Article 8(1) of the EU Charter of Fundamental Rights. As one principle states, “Privacy should not be viewed as static: it is flexible enough to adapt and develop […] in light of rapidly changing technological and social conditions.”
Equality and non-discrimination form another cornerstone. AI’s reliance on rules rather than individual assessment can embed bias from training data, design choices, or deployment contexts. Documented cases range from gender bias in job advertising, racial bias in recidivism prediction tools, and healthcare algorithms under-referring black patients, to recruitment systems discriminating against women. Human rights law prohibits direct, indirect, and structural discrimination, requiring fairness and due process. As noted, “Adopting well-established and internationally accepted standards in human rights law minimizes the need for fresh debates on highly contested concepts in ethics.” Effective mitigation demands diverse development teams, bias detection tools, and regular algorithmic audits to prevent proxies like postcode from serving as stand-ins for protected characteristics.
Autonomy faces new threats from empathic AI—systems capable of detecting, interpreting, and simulating human emotions. Applications range from driver drowsiness detection to assistive technologies, but risks include surveillance, behavioral chilling effects, and manipulation. Emotion recognition in workplaces or schools can infringe privacy and freedoms of thought, expression, and association. Manipulative uses blur lines between influence and coercion, especially in political disinformation campaigns. The draft EU AI Act seeks to prohibit subliminal techniques causing harm and restrict certain trustworthiness profiling. Children’s rights are particularly sensitive; the UN Committee on the Rights of the Child has called for bans on emotional analytics targeting minors.
Economic and social rights—education, health, social security, work—also intersect with AI. Properly designed systems can advance Sustainable Development Goals, but equitable access is essential to avoid deepening social divides. Research and funding should prioritize technologies benefiting all communities, not solely those yielding high profits.
Fairness and due process in AI decision-making require transparency, human oversight, and avenues for challenge. Decisions affecting rights must disclose the decision-maker, criteria used, and allow verification of data accuracy. Without such safeguards, individuals risk being subjected to opaque, unreviewable outcomes—the proverbial “computer says no.” Human rights law’s procedural fairness standards can guide minimum requirements for transparency and accountability.
AI’s reach extends to other rights: freedom of expression and information in content moderation, family life in child safeguarding analytics, freedom of assembly in facial recognition, and even the right to life in military applications. Across these domains, existing human rights norms offer tested boundaries to define acceptable uses and prevent harm.