May 14, 2026

Author: Lawyer Enrico Germano

2. Corporate compliance and de facto regulation

In the absence of specific regulations, AI-related risk management has so far been largely entrusted to corporate compliance. Companies are required to integrate AI into their internal control systems, leveraging general due diligence obligations and cross-sector regulations, including the Federal Act on Data Protection-FADP.

We are thus observing the emergence of practices such as algorithmic audits, documentation of decision-making processes, and the introduction of human oversight mechanisms. These tools certainly help mitigate risks, but they have a clear limitation: their adoption is driven more by reputational and economic considerations than by clear and uniform legal obligations.

This form of “private regulation” produces uneven effects and does not guarantee an adequate level of protection. Furthermore, it implicitly shifts onto businesses the responsibility of filling regulatory gaps that would otherwise require legislative intervention.

3. Towards an Independent Liability Regime for AI

In light of these critical issues, it is necessary to consider the need to introduce a specific liability regime for artificial intelligence in Swiss law.

Such a regime should first take into account the evidentiary difficulties inherent in damages caused by algorithmic systems. From this perspective, an initial approach could consist of mitigating or reversing the burden of proof in favor of the injured party, at least in cases where the opaqueness of the system makes it impossible to reconstruct the decision-making process.

Secondly, it seems appropriate to consider introducing forms of strict liability for certain categories of high-risk systems, along the lines of those established in other areas characterized by intrinsic danger.

Another key element concerns transparency and traceability obligations. Imposing minimum standards of documentation and explainability would not only facilitate the determination of liability but would also help strengthen trust in the technologies used.

Finally, an autonomous regime should explicitly address the issue of the plurality of actors involved, providing clear criteria for assigning responsibility along the technology supply chain.

The goal is not to completely replicate foreign models, such as the European Union’s AI Act, but to develop a solution consistent with the principles of Swiss law, while also being responsive to international developments.

Conclusion

Artificial intelligence challenges fundamental assumptions of tort law, revealing the limitations of a system built around the centrality of human action.

In the Swiss context, the lack of specific regulation has so far favored a pragmatic approach, based on the adaptation of existing rules and the centrality of corporate compliance.

However, this approach is now showing its limits. The difficulty of proving causality, the fragmentation of liability, and the opaqueness of algorithmic systems risk compromising the effectiveness of legal protection and generating uncertainty for economic operators.

In Swiss law, you can start from the principles of liability for tort (Art. 41 Federal Act on the Amendment of the Swiss Civil Code-Part five: Code of Obligations), liability of the owner of installations/danger, or contractual liability. The central problem is that AI breaks the classic perpetrator → action → damage link.

Indeed, it is clear that there are regulatory gaps in Switzerland, characterized by the absence of specific regulations. By comparison, unlike the European AI Act, there is no risk classification, no uniform transparency requirement, and no specific strict liability.

Therefore, significant open issues remain in the sensitive areas of proving damage and causation, with excessively high evidentiary standards, the attribution of liability in unregulated technological sectors, and in the context of automated decisions, with still limited protection.

Switzerland risks a double disadvantage: internal legal uncertainty and misalignment with European law or other foreign countries.

In this scenario, the introduction of an autonomous liability regime for artificial intelligence is not a merely theoretical option, but a systemic necessity.

Only through targeted regulatory intervention will it be possible to re-establish a balance between technological innovation and the protection of rights, while also ensuring legal certainty and the competitiveness of the Swiss legal system in the international context.