Text-drafting, code-generating and analytical AI tools are now part of many professional workflows – from the accounting office through the IT developer to the consultant. When AI-assisted work gets something wrong, the question of liability arises just as it would with any other professional mistake – only the picture has one extra layer of complexity. This article unpacks that layer.
- Using an AI tool does not, on its own, remove the professional's liability: for the work you deliver, you or your business will typically still be responsible.
- Professional indemnity insurance typically looks at the activity, not the tool used to carry it out – but using a new tool can amount to a material change of activity.
- The typical forms of an AI error – incorrect data, a flawed calculation, a fabricated reference, copyright risk – fit within the traditional categories of professional error.
- Because of how claims-made cover works, it matters that your use of AI isn't left out of the conversation with your insurer.
Who is liable if the AI gets it wrong?
In factual, informational terms: under current market and legal practice, an AI tool is treated as an aid, and the liability of the professional or business delivering the final output does not disappear simply because part of the work was done by a machine. Just as citing a poor source or a flawed piece of software-based calculation doesn't excuse a professional from liability, an error made by an AI tool doesn't turn it into "no one's territory" either. The precise legal assessment depends on the case, on the applicable laws, and on case law as it develops – this article flags that, it doesn't settle it.
Does this change your declared activity?
If the AI tool is only background support – faster drafting, putting together an outline, say – your declared activity typically doesn't change in any material way. It's a different matter if an AI-based solution becomes a standalone part of your service: for example, a consultant starts selling an AI-based analysis module, or a developer passes on AI-generated code without quality assurance. Changes of that kind are worth sharing with your insurer, following the logic set out in the reporting activity changes article – staying quiet about it is, over time, a bigger risk than raising it.
Why is this especially sensitive for claims-made products?
Professional indemnity insurance typically works on claims-made logic: what matters is when a claim is made against you (and when it is notified to the insurer), and what already counted as a "known circumstance" when the policy was taken out. If an AI-assisted workflow has already been running for some time and a suspected error emerges along the way, the timing of notification and the question of the retroactive date – as the claims-made and retroactive date article sets out in detail – isn't a theoretical point: it can determine whether a claim that surfaces later falls within cover.
Typical risk points by profession
For an IT developer, the classic risk is a hidden bug or a security gap in AI-generated code – the IT developer and consultant article covers this in more depth. For an accountant or a consultant, incorrect data in a summary that was put together quickly with an AI tool but not properly checked can cause a loss. For copywriters and marketers, the exposure lies in copyright risk and in carrying over a factual claim "invented" by an AI that doesn't actually exist. For trainers, the source can be an inaccuracy in AI-based course material. The common thread: the loss doesn't come from the tool itself, but from how it's built into the workflow.
What can you do in practice?
A few simple habits count for a lot: always have a human check the AI tool's output before it reaches the client; it's worth verifying the more important findings and sources separately, especially where they contain figures or references to legislation; and if AI use becomes a material part of the service, it's worth raising explicitly both when requesting a quote and when reviewing your policy. Documentation – when an AI tool was used, for what purpose, and with what checks – is just as valuable in a dispute as documentation for any other professional workflow.
Common question: "Do I need separate cover for this?"
A standalone insurance product specifically designed for "AI risk" is not yet common on the Hungarian market – this may change over time, as the risk and claims data mature. What's a realistic step now is to explicitly raise your use of AI tools when requesting a quote for, or reviewing, your existing professional indemnity cover, so that you and your insurer are looking at the same picture of your activity. This isn't primarily a question of extra cost – it's a question of accuracy, and it's exactly what a 20-minute conversation can clarify.
A question tailored to your own trade? Book a free 20-minute consultation →
