How AI and automation is driving liability

In the last few years, AI and automation have spread everywhere – including into our day-to-day work. It’s become a useful tool for businesses, helping to draft reports, analyse data, make decisions and interact with clients. But there’s a downside – according to a 2024 report by Clyde & Co, 72% of UK professional indemnity market respondents say that AI is increasing professional risk exposure¹.

The truth is that AI can make mistakes too. But when it makes a mistake, who’s responsible for it? AI can’t face consequences because it’s a tool, not a person.

Simply, AI requires oversight from the business, so it’s the business that is held responsible for any errors that might have slipped through the cracks. If an AI produces content for the business and that content isn’t double-checked and reviewed by someone, it’s the fault of the business. The liability continues to sit with the business, so saying that the AI made the mistake won’t cut it.

AI hallucinations

AI can seem like magic, seemingly producing answers out of thin air to our queries, or generating pages of content from data sets in a matter of seconds. But unfortunately, sometimes that’s actually the truth – they can occasionally produce answers out of thin air, with no basis in reality at all. This is called an AI hallucination, and it happens when the AI chatbot confidently answers a question with a statement that is completely wrong and seems to be entirely made up. So, it’s important to double-check every answer that the chatbot gives you before using it in content for the business².

Bias and data issues

Businesses also need to stay aware of the possibility of bias and data issues from AI. If an AI model is trained on data that’s unbalanced or incorrect, it won’t necessarily flag the issue but will simply give results based on that data, meaning that the results could be unfair or even discriminatory. Data should be double-checked before being fed to an AI, and the same goes for the results given. Don’t just assume that the AI is always right.

Lack of explainability

There’s also a lack of explainability that comes with using AI. We often don’t understand how an AI system comes to a conclusion or makes a prediction – and this can lead to a headache for businesses if the AI makes a mistake. Strict data governance laws (like GDPR) say that people have "the right to an explanation" for automated decisions. If a business fails to provide this, it can lead to severe penalties^3/sup>.

The lack of explainability goes even further. You may find that stakeholders and customers are uncomfortable about automated outputs if they can’t see the reasoning behind them, and without the visibility into AI reasoning, your business may inherit hidden biases which were embedded in the training data. And if your AI chatbot hallucinates or makes a bad decision, it can be difficult for your software developers and engineers to find out why.

The information contained in this article is for general information purposes only. It does not constitute legal or other professional advice and cannot be relied upon as such. Should you have any queries, we recommend that you consult the appropriate professional adviser. The links provided in this document are for reference only. Please note that we are not responsible for the content of any linked site.

Sources:

1. https://www.clydeco.com/en/reports/2024/11/professional-indemnity-market-report-2024

2. https://www.fsb.org.uk/resources/article/ai-risks-for-businesses-5-practical-ways-to-innovate-safely-MCHKMECWKHEZBIBFFYTTWYV23ZAY

3. https://gdpr-info.eu/art-22-gdpr