As artificial intelligence (AI) continues to permeate various sectors, the debate around its ethical deployment intensifies. A new study published in AI & Society argues that the prevailing focus on trustworthiness as a criterion for AI governance may overlook a crucial aspect: the human vulnerabilities that arise from reliance on automated systems. The research contends that understanding these vulnerabilities is essential to building trustworthy AI.
The paper, titled “The value of vulnerability for trustworthy AI,” shifts the narrative around trustworthiness from a mere branding exercise to a framework for accountability among developers, deployers, and regulators. It poses the question of whether AI governance is genuinely designed to protect those most affected by its deployment, rather than just fostering public confidence for wider adoption.
Trustworthiness as a concept has gained traction in global AI policy, notably highlighted by guidelines from the European Commission’s High-Level Expert Group. These guidelines present trustworthiness as the ethical goal of AI development, a sentiment echoed in G20 policy statements and various governance documents. While this focus aims to balance the enthusiasm for AI’s benefits with concerns about social and ethical risks, the study argues that it has rendered the term “trustworthiness” vague and undefined.
The study critiques the instrumental nature of trust, arguing that it is often seen as a prerequisite for adopting AI systems rather than as a deeper social good. This approach risks reducing ethical considerations to a checklist of values, overlooking the need for a more profound understanding of how vulnerability shapes human experience in relation to AI. Trust becomes a means to achieve cooperation and stability, but the study warns that framing trust as the ultimate goal may lead to ethical frameworks that lack meaningful substance.
Several criticisms emerge within the study regarding the notion of trustworthy AI. For instance, can AI systems be regarded as trustworthy in a moral sense? Traditional human trust entails expectations of goodwill and responsibility, qualities that technical systems do not possess inherently. Moreover, the focus on trustworthiness can obscure accountability, leaving unclear whether the responsibility lies with the technology, its developers, or the broader socio-technical ecosystem.
The paper also addresses the risk of ethical dilution, suggesting that trustworthiness can serve as a façade for ethical complexities, allowing organizations to sidestep serious discussions about power dynamics and regulation. In this context, the study advocates for a redefinition of trustworthiness through the lens of vulnerability. Instead of merely evaluating whether AI appears trustworthy based on established criteria, stakeholders should consider the vulnerabilities that compel individuals to rely on AI and the new vulnerabilities introduced by these systems.
Vulnerability, according to the study, is a fundamental aspect of human existence rather than a concern limited to specific at-risk populations. Trust emerges as a mechanism to navigate vulnerability, allowing individuals to engage with others and systems without constant oversight. However, placing trust in someone or something inherently introduces new risks, as it exposes individuals to potential misuse or neglect.
The study emphasizes that discussions on vulnerability in AI often focus narrowly on certain groups needing protection, overlooking the broader impact of AI systems in creating new vulnerabilities. AI technologies increasingly shape environments where individuals operate, influencing decisions in areas like healthcare, policing, and employment. Therefore, the paper argues, it is critical to recognize how AI can not only perpetuate existing inequalities but also restructure the very conditions of dependence in society.
By centering vulnerability in the conversation around trustworthy AI, the study redefines the concept as a sociotechnical undertaking aimed at recognizing and addressing the vulnerabilities of all stakeholders involved, from development to deployment. This approach necessitates participatory governance, wherein those affected by AI systems are actively involved in their design and oversight. Such participation can illuminate power imbalances and early harms, fostering accountability and trust.
The author argues for a shift in the understanding of digital and technological sovereignty. Rather than merely managing the balance between innovation and rights, legitimate authority over AI should prioritize protecting individuals whose lives are increasingly influenced by these technologies. This calls for governments to not only certify acceptable risks but to actively identify and mitigate the vulnerabilities that arise from AI dependencies.
While principles such as safety, accountability, transparency, and fairness remain vital, they should be viewed as tools that derive their significance from a foundational commitment to understanding vulnerability. Without this commitment, ethical guidelines risk becoming detached from the essential questions of accountability and the purpose of AI governance. Ultimately, the study lays bare a pressing query: Are current AI systems being developed and regulated in ways that genuinely justify public trust?
See also
OpenAI’s Rogue AI Safeguards: Decoding the 2025 Safety Revolution
US AI Developments in 2025 Set Stage for 2026 Compliance Challenges and Strategies
Trump Drafts Executive Order to Block State AI Regulations, Centralizing Authority Under Federal Control
California Court Rules AI Misuse Heightens Lawyer’s Responsibilities in Noland Case
Policymakers Urged to Establish Comprehensive Regulations for AI in Mental Health
