AI Liability and Accountability: Who is Responsible When AI Makes a Harmful Decision?

By Soham NandiReviewed by Bethan DaviesFeb 19 2025

Artificial Intelligence (AI) is becoming a crucial tool in major fields such as healthcare and transportation. However, as AI systems become more autonomous and influential in decision-making, concerns about AI-related harm are growing.

Image Credit: Teerachai Jampanak/Shutterstock.com

High-profile incidents, such as self-driving car accidents and AI-driven medical misdiagnoses, have underscored the serious consequences of AI failures. These cases raise a critical question: when AI causes harm, who should be held accountable—the user, the developer, or the AI system itself?^1,2

Traditional liability systems often struggle to address the unique challenges posed by AI-related cases. This growing debate underscores the need for clearer frameworks to assign responsibility, ensure justice for victims, and avoid discouraging AI innovation.¹

This article will explore potential harm caused by AI, evaluate how existing legal frameworks respond to such cases, and discuss how accountability can be better addressed in the legal landscape.

AI laws are changing fast—Don’t get left behind. Download your free PDF now!

Understanding AI Liability

AI liability refers to the legal responsibility for damages or harm caused by artificial intelligence systems. As AI technology becomes increasingly integrated into various aspects of life, governments and legal systems worldwide are grappling with how to address the unique challenges posed by AI-related disputes.

One of the biggest hurdles in establishing AI liability is the complexity and opacity of these systems. The "black box" nature of AI makes it difficult to determine how harm occurred, complicating traditional legal concepts such as breach, defect, and causation. Additionally, liability may involve multiple parties—including developers, deployers, and users—making it even harder to assign responsibility.

To address these challenges, some jurisdictions are considering measures to ease the burden on claimants. These include presumptions of causality and requirements for AI providers to disclose relevant evidence. Many proposed frameworks also allow non-contractual, fault-based claims for AI-related harm, complementing existing strict liability regimes. Some regulations even have extraterritorial reach, applying to AI systems operating within specific markets, regardless of the provider’s location.

While some have proposed the idea of AI personhood, current legal frameworks generally do not recognize AI as a legal entity. Instead, liability remains with the human actors involved in AI development, deployment, and use. As AI technology evolves, lawmakers and regulators worldwide are working to strike a balance between consumer protection and innovation. The goal is to ensure that victims of AI-related harm have access to compensation while providing legal certainty for businesses investing in AI.^1,4

AI-Related Harm and Accountability Issues

There have been several cases highlighting the challenges surrounding AI liability and accountability. One notable example is the series of self-driving car accidents, particularly the fatal crash involving an Uber autonomous vehicle in 2018. In this incident, the vehicle, which was equipped with AI software, failed to recognize a pedestrian, leading to a tragic accident. This incident led to the question, who is responsible for this harm—Uber, the vehicle’s manufacturer, or the AI system itself? The case exposed significant gaps in existing liability frameworks, as autonomous vehicles operate in complex, unpredictable environments that defy simple attribution of fault.^4,5

In healthcare, AI systems used for diagnosing diseases have also been involved in controversy. A case in which an AI system misdiagnosed a patient's condition, leading to a delay in treatment, raised similar concerns. While AI algorithms are designed to assist medical professionals, when mistakes occur, it can be unclear whether liability falls on the developers of the AI, the healthcare providers, or both.^1,6

These cases highlight the difficulty of applying traditional legal frameworks to AI systems, where accountability often lacks clarity. They expose the need for updated liability systems that can handle the nuances of AI-driven decisions, ensuring fairness and protection for those affected by AI-related harm.^1,4

So, who should be held accountable for AI decisions?

Frameworks for Assigning Accountability in AI-Related Harm

Although the recent developments surrounding AI have been positive in many regards, its rapid advancement has somewhat outpaced the development of robust accountability frameworks, creating a complex landscape of potential approaches—each with its own strengths and limitations.

Traditional models, such as the principal-agent model (similar to the doctor-medical student relationship), place liability on the professional overseeing the AI system. While straightforward, this approach may discourage AI adoption, as practitioners may be reluctant to take responsibility for AI failures they cannot fully control or understand.

Another common approach, the product liability model, allows any entity in the AI supply chain to be held accountable. However, AI’s inherent unpredictability and the "black box" nature of many systems make it difficult to prove specific defects, limiting this model’s effectiveness.

Emerging frameworks are seeking alternative solutions for AI accountability. The reconciliation-based approach aims to distribute responsibility equitably among developers, users, and consumers, balancing liability, victim compensation, and preventive measures. Another proposal, the Computational Reflective Equilibrium (CRE) framework, takes a context-driven approach by aligning accountability with each party’s level of control over the AI's actions. This adaptable model considers the technical complexity and unpredictability of AI systems to ensure fairness in assigning responsibility.^1-3

The broader debate on AI accountability also raises ethical concerns, such as balancing transparency with intellectual property protections, individual rights with societal benefits, and innovation with risk mitigation. While AI developers often advocate for limited liability to encourage innovation, consumer rights groups push for stronger protections and clearer accountability, leaving policymakers to navigate these competing interests.

As AI technology continues to evolve, accountability frameworks must adapt. Key challenges include addressing the increasing autonomy of AI systems, developing effective methods to audit and explain AI decision-making, and establishing international standards for AI accountability. The future may see hybrid models that combine elements from multiple frameworks tailored to specific AI applications and risk levels.^1-3

Global Approaches to AI Accountability

Governments around the world are working to develop legal responses to help mitigate some of the associated risks of AI. However, the strategies and regulatory approaches vary significantly across different regions:^7-9

• European Union (EU): The proposed AI Act and AI Liability Directive aim to close accountability gaps by shifting some burden of proof onto developers and operators. This legislation is designed to make it easier for victims of AI-related harm to seek compensation and hold developers accountable for failures.
• United States (US): The government has issued an executive order on Safe, Secure, and Trustworthy AI, emphasizing responsible AI development while leaving detailed liability issues to state and industry-specific regulations.
• ASEAN Nations: In 2024, ASEAN endorsed the AI Governance and Ethics Guide, providing guidelines for responsible AI deployment across member states while encouraging innovation.
• Middle East: The UAE and Saudi Arabia have adopted a "soft-law" approach with AI ethics guidelines, focusing on flexible, principle-based regulation to foster AI growth while managing risks.
• India: While India lacks AI-specific legislation, regulators support a risk-based approach, particularly in sectors like healthcare. The Indian Council of Medical Research has issued ethical AI guidelines emphasizing developer accountability for AI-driven decisions.
• United Kingdom (UK): The UK has signed the first international treaty addressing AI risks, reinforcing its commitment to preventing AI-related harm while promoting responsible AI development.

So, How Can You Prevent AI Liability?

Avoiding AI liability isn’t just about following the rules—it’s about being proactive. The best way to protect yourself (and your company) is to focus on transparency, thorough testing, and solid documentation.

One major challenge in AI-related legal cases is that it’s often tough for victims to prove harm. To stay ahead of potential issues, keep clear records of how your AI system is designed, what data it uses, and how it makes decisions. If a legal issue ever comes up, having this documentation can show that you did your due diligence.

If this was valuable to you, share it with your network and keep the conversation going!

Testing is another big piece of the puzzle. Regularly checking your AI system for errors and biases can help catch problems before they become liabilities. Using diverse, representative data is key to avoiding discrimination claims, which can be a major legal headache.

It’s also smart to have an internal accountability framework—basically, a clear plan for who’s responsible for compliance and risk management. Laws around AI are changing fast (think the EU’s AI Act and Product Liability Directive), so staying updated is crucial. And for extra protection, consider getting insurance designed for AI-related risks.^1,3,4

Taking these steps now can save you a lot of trouble down the road, keeping your AI both legally compliant and trustworthy.

The Future of AI Liability

As AI technology advances, liability frameworks will continue evolving. The EU’s AI Liability Directive suggests a future where the burden of proof shifts more toward developers. Greater transparency, documentation, and risk assessments will likely become standard requirements.¹

As AI becomes more autonomous and embedded in critical sectors, regulatory approaches must balance innovation with consumer protection. Companies should anticipate stricter legal standards and proactively align their practices with emerging global regulations.⁴ A refined approach to AI liability will be crucial in ensuring fairness and justice and fostering continued innovation.

Want to Learn More?

If you're interested in exploring more about AI liability and its broader implications, why not check out the articles below?

• Sustainable AI: Balancing Tech and Ethics
• The Pro-Innovation AI Regulations in the UK: What You Need to Know
• Navigating the EU AI Act: Implications for the Robotics Industry
• What Does the EU AI Act Mean for Artificial Intelligence Regulation?
• How the United States AI Regulatory Landscape Differs from the EU and UK
• Ethical Considerations in AI-Driven Healthcare

Stay tuned for our next article, where we’ll explore "How AI and Digital Transformation Are Shaping India’s Regulatory Landscape." We’ll be looking into the latest policy developments, government initiatives, and the impact of AI on India’s evolving regulatory framework.

References for Further Reading

1. Bottomley, D., & Donrich Thaldar. (2023). Liability for harm caused by AI in healthcare: an overview of the core legal concepts. Frontiers in Pharmacology, 14 DOI:10.3389/fphar.2023.1297353
   https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2023.1297353/full
2. Smith, H. (2020). Clinical AI: opacity, accountability, Responsibility and Liability. AI & Society, 36. DOI:10.1007/s00146-020-01019-6
   https://link.springer.com/article/10.1007/s00146-020-01019-6#Sec8
3. Ge, Y., & Zhu, Q. (2024). Attributing Responsibility in AI-Induced Incidents: A Computational Reflective Equilibrium Framework for Accountability. ArXiv.org. DOI:10.48550/arXiv.2404.16957
   ‌https://arxiv.org/pdf/2404.16957
4. Goudkamp, J. (2023). Automated Vehicle Liability and AI. Social Science Research Network. DOI:10.2139/ssrn.4509872
   https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4509872
5. Lee, D. (2019, November 20). Uber's self-driving crash “mostly caused by human error.” BBC News. Available at: https://www.bbc.com/news/technology-50484172 (Accessed on 18^th February 2025)
6. Saenger, J. A., Hunger, J., Boss, A., & Richter, J. (2024). Delayed diagnosis of a transient ischemic attack caused by ChatGPT. Wiener Klinische Wochenschrift, 136(7-8), 236–238. DOI:10.1007/s00508-024-02329-1
   https://link.springer.com/article/10.1007/s00508-024-02329-1
7. Ministry of Justice. (2024, September 4). UK signs first international treaty addressing risks of artificial intelligence. GOV.UK. Available at: https://www.gov.uk/government/news/uk-signs-first-international-treaty-addressing-risks-of-artificial-intelligence (Accessed on 18^th February 2025)
8. York, U. of. (2024). AI regulation and policy landscape in the Middle East. University of York. Available at: https://www.york.ac.uk/assuring-autonomy/news/blog/ai-regulation-middle-east/ (Accessed on 18^th February 2025)
9. CRAFTING A LIABILITY REGIME FOR AI SYSTEMS IN INDIA Crafting a Liability Regime for AI Systems in India. (2024). Available at: https://aiknowledgeconsortium.com/wp-content/uploads/2024/10/ReportESYACentreReport-CraftingaLiabilityRegimeforAISystemsinIndia.pdf (Accessed on 18^th February 2025)

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.