Should AI Liability Insurance Be Mandatory? EU, Korea and Türkiye

Correction (19 October 2026): This article originally presented the EU AI Liability Directive (AILD) as a live draft with a presumption of causality; the Commission withdrew that proposal in 2025, and the EU section now describes the revised Product Liability Directive instead.

If an autonomous delivery drone spirals out of control and injures a pedestrian, who will pay the hospital bills? When an algorithmic stock market bot managing a multi-million-dollar portfolio experiences an "AI hallucination" and suddenly wipes out the funds, who compensates the loss? Risks like these have opened a new legal debate worldwide: Should Artificial Intelligence Financial Liability Insurance be mandatory?

This article examines the problem of compensating harm caused by autonomous systems. It compares the legal systems of the European Union (EU), South Korea (KR), and Turkey (TR), three jurisdictions where technological automation is most intense, under four headings: "Background," "Regional Regulations," "Implications," and "Conclusions."

1. Background: The Evolution of Liability Law and the "Black Box"

In traditional tort law, the general rule is that damages are payable only on the basis of "Fault-based Liability". Modern AI systems (especially Deep Learning models), however, work as a "Black Box" that sets its own rules. Even the engineer who built the system cannot fully explain why the machine preferred data in the 45th layer over data in another.

1.1. The Collapse of the Burden of Proof

If an autonomous vehicle hits a pedestrian, the classical rules require the pedestrian to prove that "the software developer made a coding error" or "the manufacturer installed the sensors incorrectly." An ordinary person has practically no chance of proving in court a technical flaw in the closed-source algorithm of a billion-dollar tech company. In practice, the tech companies would never pay, and the whole loss would fall on the victim.

1.2. The Solution: "Strict Liability" and an Insurance Pool

The law developed "Strict (Hazard) Liability" precisely to overcome this kind of evidential impossibility. Operators of nuclear power plants or transporters of explosives are held liable for damages even if they are not at fault. Legal scholars now argue that high-risk AI belongs in this "hazardous activity" category and that a Mandatory AI Insurance pool should be set up, much like Mandatory Traffic Insurance.

2. Global Axis Analysis: European Union, South Korea, and Turkey

Against this background, let us look at how some of the world's leading technology and legal jurisdictions approach the problem.

2.1. European Union (EU): The Revised Product Liability Directive (PLD) and the Withdrawn AILD

Besides enacting the world's most comprehensive Artificial Intelligence Act (EU AI Act), the European Union adopted a revised Product Liability Directive (PLD) in 2024 (Directive (EU) 2024/2853), which expressly covers software, including AI systems. A separate Artificial Intelligence Liability Directive (AILD) was proposed in 2022, but the Commission announced its withdrawal in February 2025 and formally withdrew it later that year.

• Easing the Burden of Proof: Recognizing that victims cannot be expected to read code, the revised PLD allows national courts to presume that a product was defective, or that the defect caused the damage, where technical or scientific complexity makes proof excessively difficult for the claimant and the claimant shows that this is likely. The presumption is rebuttable: the manufacturer can still prove otherwise. (The AILD's separate "presumption of causality" for AI systems fell away with its withdrawal.)
• Insurance Status: The current texts do not yet impose a blanket "Mandatory Insurance" requirement. In practice, however, commercial insurance is becoming de facto mandatory for companies releasing high-risk AI (e.g., autonomous robots, medical AI), given the scale of potential fines and compensation claims. The European Parliament called as early as October 2020 for compulsory insurance for operators of high-risk AI systems, but no such obligation has been adopted.

2.2. South Korea (KR): Robotic Insurance and Proactive Tech Law

South Korea is one of the most highly automated countries in the world, particularly in manufacturing robots, service bots, and medical AI. Unlike the EU's "rule-maker" approach, Korea favors specific, sector-by-sector solutions so as not to stifle innovation.

2.3. Turkey (TR): The Current Vacuum and the Potential of the Code of Obligations

Turkey currently has no specific law on "Artificial Intelligence" or "Autonomous Systems." Where harm occurs, the Turkish Code of Obligations (TBK) No. 6098 applies. However, the TBK's traditional rules are ill-suited to complex harm caused by algorithms (for instance, compensating a company that unjustly lost a multi-million lira tender due to algorithmic discrimination).

The strongest argument currently available in Turkey is "Strict (Hazard) Liability" regulated under TBK Article 71. If an activity has the inherent potential to cause severe damage to the environment (explosives factory, hazardous material transport), the operator is held strictly liable. Just as Medical Malpractice or Highway Traffic Insurance is mandatory in Turkey, legal scholars strongly recommend making "Autonomous Decision Support Systems Financial Liability Insurance" mandatory under the upcoming Turkey AI Action Plan / AI Act. Otherwise, foreign AI companies operating in the Turkish market will avoid paying for the harm they cause (by pointing out that "you agreed to the user terms").

3. Implications: InsurTech and Risks for Corporations

If states make insurance mandatory for high-risk AI systems, the effects on the technology and financial sectors will be considerable:

1. Actuarial Crisis (Impossibility of Premium Calculation): An insurer pricing traffic insurance can draw on 100 years of accident data. Nobody knows when a self-coding AI will make a mistake. How is the premium for an opaque "Black Box" algorithm to be calculated? Insurers may respond by demanding very high risk premiums, driving up the cost of AI sharply.
2. Stifling of Innovation: Large corporations such as Samsung and Hyundai, as the South Korean example shows, can easily afford high premiums for their autonomous robots. But if a promising AI start-up founded by 3 recent graduates has to pay millions of liras in "mandatory financial liability insurance" to launch its product, innovation will remain solely in the hands of global monopolies.
3. The Rise of InsurTech: To address exactly this problem, "InsurTech" (insurance technology), which uses AI itself to measure the risks of algorithms before policies are issued, will grow quickly. A new generation of auditing firms will emerge to test an AI company's code against millions of scenarios in a virtual lab (Red-Teaming) before a policy is issued.

4. Conclusions: What Awaits Us in the Future?

When automobiles first appeared, they were governed by the rules for horse-drawn carriages, until rising numbers of fatal accidents brought "Mandatory Traffic Insurance" into being. AI is now going through the same painful process. Both the European Union's legislative framework and South Korea's practical measures show that AI insurance will inevitably become mandatory in high-risk areas.

Turkey should not import this technology while leaving the risks on the shoulders of its own citizens. The legal framework it adopts should protect citizens through a centralized "AI Damage Compensation Pool," similar to medical malpractice or traffic insurance, while subsidizing premiums to a level that does not stifle entrepreneurs.