Artificial intelligence is taking on an increasing number of tasks within companies: it analyses markets, evaluates figures and prepares decisions. However, if an AI-based analysis is incorrect and the company suffers a loss as a result, a very practical question arises: who is liable?
The answer is clear: the Executive Board
No matter how sophisticated the AI system used may be, legal responsibility for business decisions remains with the governing body. This applies equally to the board of directors of a public limited company and to the managing directors of a limited liability company. AI can assist in the preparation of decisions, but it cannot develop independent judgement nor contribute company-specific experience.
However, this is precisely what is crucial when it comes to the so-called ‘business judgement rule’. This provision under company law protects board members from personal liability if they have taken a business decision in good faith, on the basis of adequate information and in the interests of the company. However, anyone who simply accepts AI results without scrutinising them critically and applying their own judgement cannot rely on this protection.
AI is no substitute for an expert opinion
Another misconception is becoming widespread in practice: many managers treat AI analyses as if they were expert reports from an external specialist. This is legally problematic. An expert opinion only exonerates a governing body if the expert is genuinely competent, has not been given any result-oriented instructions and was fully aware of all relevant information. As things stand, AI cannot meet these requirements. An erroneous AI result therefore does not have the same legal implications in terms of liability as a human expert opinion.
What must be documented
Anyone who incorporates AI into decision-making processes should carefully document its use. It must be possible to ascertain, in retrospect, on what information a decision was based. The documentation should include, in particular:
- Name of the person who used the AI, and the date of use
- Name and version of the AI tool used
- Inputs and prompts used, as well as uploaded documents
- Checking the results for context, timeliness and sources
- Compliance with internal approval processes
Prompts and results should be saved as PDF files wherever possible, as many AI tools do not offer permanent storage.
The EU AI Act introduces additional obligations
The EU AI Act classifies certain AI applications as high-risk and requires, amongst other things, a risk management system that must be kept up to date, as well as strict documentation requirements. Anyone who ignores these regulatory requirements is in breach of the so-called duty of legality and thereby exposes themselves to an additional liability risk. AI is therefore not just a technological issue, but a genuine compliance issue.
What this means for managing directors of limited liability companies (GmbHs)
Although the discussion often relates to the executive boards of public limited companies, the principles apply equally to managing directors of limited liability companies (GmbHs). Anyone who uses AI in day-to-day operations – for example, for financial analysis, HR decisions or strategic planning – should introduce clear internal rules, train staff and regularly review the use of AI. The attitude that ‘the IT department or the legal department will sort it out’ is not enough. Responsibility cannot be delegated.
Companies that are already making extensive use of AI, or are planning to do so, should clarify liability issues at an early stage. Legal advice can help them to develop appropriate processes and documentation structures.
Source: Haufe: Human Resources
This post and the accompanying image were created with the help of AI and have been editorially reviewed.