# AI liability and autonomous cars: will insurance replace tort?

URL: https://insurancelawestonia.ee/blog/2026-11-01-ai-liability-autonomous-vehicles-insurance/
Author: Dr Olavi-Jüri Luik, Partner, Head of Insurance Law, attorney at law (Advokaadibüroo WIDEN OÜ)
Published: 2026-11-01
Last updated: 2026-11-05

> Autonomous vehicles, liability gaps and motor insurance in Estonia and the EU: mandatory cyber, product liability, MID reform or no-fault? Luik and Kinkar.

Partially autonomous vehicles are already on Estonian roads, and the progressive transfer of the driving function from humans to artificial intelligence raises a basic question: when an AI-driven vehicle causes harm, who is liable and who pays? This post is based on research by Olavi-Jüri Luik and Rauno Kinkar published in Estonian in *Juridica* (2024) and in English in *Juridica International* (2025), updated as of November 2026.[^1]

## New risks, old liability rules

As autonomy increases, the cause of an accident shifts from human behaviour to software, sensors and data. Autonomous vehicles add distinct risks: software errors and algorithmic decisions that cannot be traced to any individual's negligence; cyberattacks (hacking, sensor spoofing); inaccurate data and maps; and unclear allocation of roles during handovers between system and human. Fault-based liability struggles with these risks, and product liability claims face evidential hurdles and the development-risk defence.

## Estonia's starting point is robust

Under the Law of Obligations Act (*võlaõigusseadus*, VÕS) § 1057, the possessor of a motor vehicle is strictly liable as the possessor of a major source of danger, and compulsory motor third-party liability insurance covers that liability.[^2] Since the 2024 reform, the Motor Insurance Act (*liikluskindlustuse seadus*, LKindlS) no longer contains its own liability rules; liability follows the VÕS and the motor insurer covers it.[^3] For third-party victims this is resilient to automation: because the possessor's liability does not depend on fault, it does not matter whether a human or an algorithm caused the accident. The motor insurer pays and may then seek recourse against the manufacturer.

The gaps lie elsewhere: injury to occupants (especially the possessor), the difficulty of proving recourse claims against manufacturers, and vehicles that in practice have no human possessor-driver.

## Four options assessed

**1. Mandatory cyber insurance for vehicles.** Paying traffic losses caused by hacking through a separate compulsory cover would create demarcation disputes and divergent compensation rules. We do not recommend it.

**2. Mandatory product liability insurance for manufacturers.** Channelling risk to its source is attractive, but until data show that autonomous vehicles are riskier than conventional ones, there is no logical justification for a separate compulsory scheme. Overall risk is more likely to fall, despite new specific risks.

**3. Amending the Motor Insurance Directive.** The Motor Insurance Directive (2009/103/EC, as amended by Directive (EU) 2021/2118) already harmonises compulsory motor insurance, though not civil liability itself.[^4] The most efficient solution would be to require national motor insurance systems to cover losses from the additional risks of autonomous vehicles on the same footing as conventional risks. This would amount to partial harmonisation of liability for the operation of vehicles: a significant but justified step.

**4. A new no-fault scheme.** A scheme compensating victims irrespective of liability would close gaps most completely, but at the cost of moral hazard, higher costs and weaker deterrence, and it is hard to build without harmonised tort law in the EU.

Our answer to the title question is a cautious yes: the greater the unpredictability of risks, the more loss allocation shifts from tort law to insurance. Motor insurance already illustrates this, and AI accelerates the trend.

## Developments since 2024

- **AI Act.** Regulation (EU) 2024/1689 sets risk-based requirements for AI systems but does not regulate civil liability.[^5]
- **New Product Liability Directive.** Directive (EU) 2024/2853 replaces the 1985 directive and expressly treats software, including AI systems, as a product, with disclosure and presumption mechanisms that ease the claimant's burden. Member States must transpose it by December 2026.[^6] This will facilitate recourse claims by motor insurers against manufacturers, although the development-risk defence remains.
- **AI Liability Directive.** The European Commission announced in 2025 that it would withdraw its proposal, leaving fault-based AI liability to national law.

## Implications for insurers

- **Data as evidence.** Event data from autonomous systems will be decisive in claims and recourse; insurers should plan how to obtain and preserve it lawfully.
- **Recourse against manufacturers** will become routine and technically complex, often cross-border.
- **Accumulation risk.** A single software defect may manifest simultaneously across fleets in several countries, a risk profile closer to cyber than to traditional motor, with implications for reinsurance.
- **Own-damage covers.** Motor own-damage wordings should state clearly whether software failures, updates and cyber incidents are covered.

## Open questions

Two issues remain unresolved even after the new EU instruments: compensation for occupants, particularly the possessor, and identification of the "possessor" for vehicles without a human driver (robotaxis, autonomous freight). In practice the fleet operator is the likely candidate, but Estonian law does not yet address autonomous vehicles specifically.

## A note on terminology and structure

The research distinguishes between automation levels. At the lower levels, driver-assistance systems support a human who remains responsible and must monitor the vehicle constantly. At the higher levels, the system performs the entire driving task in defined conditions or everywhere, and the human becomes a passenger. Liability analysis differs accordingly. At the lower levels, the driver's fault remains central, and the possessor's strict liability already covers the victim. At the higher levels, the human contribution disappears, and the questions become: is the vehicle defective, who controls software updates, and who is the "possessor" in the legal sense?

For insurers, these distinctions should be reflected in underwriting questions (which systems are installed and activated), in claims investigation (was the system engaged at the time of the accident) and in recourse strategy (manufacturer, software supplier, map data provider or fleet operator). Estonian courts have not yet decided an autonomous vehicle case, but the existing strict liability framework allows them to approach such cases without new legislation.

## Conclusion

Estonia need not rebuild its liability law to handle autonomous vehicles: strict possessor liability backed by compulsory motor insurance meets victims' core needs. At EU level, extending the Motor Insurance Directive is preferable to creating parallel compulsory covers. The broader question whether AI itself should receive legal personality is addressed in a separate *Juridica* article (Luik, Saluste, Volberg 2026).[^7]

## Related pages

- [Motor claims in Estonia for foreign insurers](https://insurancelawestonia.ee/guides/motor-claims-estonia-foreign-insurers)
- [Subrogation and recovery in Estonia](https://insurancelawestonia.ee/guides/subrogation-recovery-estonia)
- [Cyber insurance and NIS2 in Estonia](https://insurancelawestonia.ee/blog/2026-11-02-cyber-insurance-nis2-estonia)
- [AI risk insurance in Estonia](https://insurancelawestonia.ee/guides/ai-risk-insurance-estonia)
- [Publications](https://insurancelawestonia.ee/publications)

## References

[^1]: O.-J. Luik, R. Kinkar. Tehisaru ja lepinguväline vastutus: kas kindlustussüsteem asendab tulevikus deliktisüsteemi? Autonoomsete sõidukite näide. – Juridica 2024/7, pp 496–508 <https://www.juridica.ee/article.php?uri=2024_7_tehisaru_ja_lepinguv_line_vastutus_kas_kindlustuss_steem_asendab_tulevikus_deliktis_steemi_au>; O.-J. Luik, R. Kinkar. Artificial Intelligence and Non-Contractual Liability: Will the Insurance System Replace the Tort System in the Future? – Juridica International 2025 (34), pp 71–89. <https://doi.org/10.12697/JI.2025.34.06>
[^2]: Law of Obligations Act (võlaõigusseadus, VÕS) § 1057 (RT I, 20.06.2026, 18; in force 01.09.2026). <https://www.riigiteataja.ee/akt/120062026018>
[^3]: Motor Insurance Act (liikluskindlustuse seadus, LKindlS) § 8(1), §§ 23–24 (RT I, 29.06.2024, 12); liability reform RT I, 29.06.2024, 2 (in force 01.12.2024). <https://www.riigiteataja.ee/akt/LKindlS>
[^4]: Directive 2009/103/EC, as amended by Directive (EU) 2021/2118. <https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32009L0103>
[^5]: Regulation (EU) 2024/1689 (Artificial Intelligence Act). <https://eur-lex.europa.eu/eli/reg/2024/1689/oj>
[^6]: Directive (EU) 2024/2853 on liability for defective products. <https://eur-lex.europa.eu/eli/dir/2024/2853/oj>
[^7]: O.-J. Luik, K. Saluste, M. Volberg. Kas tehisarule tuleks anda õigussubjektsus? [Should AI be granted legal personality?] – Juridica 2026/4, pp 331–341. <https://www.juridica.ee/article.php?uri=2026_4_kas_tehisarule_tuleks_anda_igussubjektsus_>
