We have now entered the glorious new age of artificial intelligence where a tiny digital assistant can sit quietly in the corner of our online meetings, listen to every word, turn conversations into notes, and confidently produce minutes before anyone has even finished taking off their headset. So what could possibly go wrong? Quite a lot, as it turns out…
Imagine a meeting with ten people; everyone logs on, conversations happen, opinions are shared, decisions are made and somewhere in the background one person decides that what this gathering really needs is an invisible electronic secretary. There is no announcement, discussion or even a simple “is everyone happy with this?” Instead, a cheerful little AI bot joins the conversation like an overenthusiastic intern who nobody remembers hiring, quietly absorbing everything being said and preparing to reveal its interpretation of events afterwards.
As soon as the meeting finishes, the little AI bot produces a beautifully formatted set of notes, and at first glance they look impressive: there are headings, action items, neat summaries and everything appears wonderfully organised. It’s the sort of document that makes everyone briefly question their own note taking ability, but the problem is that appearance and accuracy are not the same thing. A document can look professional while quietly containing mistakes that nobody notices until they matter.
Then someone who has actually made the effort to read the notes spots an issue: the action list is wrong, not dramatically so, not the sort of thing that immediately causes alarms and emergency meetings, but wrong enough to make them message another colleague who then points out another error then a third and suddenly the shiny futuristic assistant starts looking less like a helpful colleague and more like an uninvited guest who ate all the food then was sick in your flowerbed…
The legal issue is not simply that artificial intelligence can make mistakes because we all have been producing inaccurate minutes since the invention of meetings. Someone has always forgotten a key point, misunderstood a decision, or somehow managed to leave a meeting with a completely different understanding of what was agreed. The difference with AI is that the process is different; it’s not just a faster way of typing notes. The system is processing information, analysing conversations, identifying patterns and creating a written interpretation of what happened.
An AI note taker captures voices, names, opinions, decisions, personal comments and confidential discussions. That information may also pass through a third-party technology provider, meaning the conversation is no longer simply contained within the room or within an organisation’s own systems. The person who presses the button has effectively decided that everyone present is happy to participate in that process, but the obvious question is whether anyone actually agreed to that decision.
Under UK data protection principles, transparency matters. People should generally understand what information is being collected, why it is being collected, how it will be used and who may have access to it. A meeting participant should not have to discover afterwards that their words have travelled through an AI system, been processed by a technology company and transformed into a permanent written record. A simple announcement at the beginning of the meeting can prevent a great deal of difficulty: “Just to let everyone know, we are using AI to create draft notes. Is everyone happy with that?” It is not complicated, but it demonstrates respect and creates openness.
The next issue is accuracy. AI-generated notes are not official minutes by magic. They are a useful first draft and the danger comes when people forget this and treat the output as though a human secretary carefully listened, understood the context and checked every sentence before putting it into circulation. A human being usually understands when something sounds wrong because they understand nuance, relationships and the background behind a conversation. They know that Bob from marketing probably did not volunteer to “lead the complete restructuring of the organisation’s communications strategy” when what he actually said was that he would review one paragraph in a document. AI does not have that instinct – it hears words, identifies patterns and predicts meaning; sometimes it gets it right more often than not it creates certain statements that nobody in the room recalls as having been said so this creates a problem when notes are circulated as though they are an agreed record. The person who sends them out cannot simply shrug afterwards and say, “Well, the AI wrote it.” That is rather like sending everyone a letter containing mistakes and blaming the printer. The printer may have produced the writing on the paper, but someone still chose to put it in the envelope. Responsibility remains with the person or organisation that decides to distribute, rely upon or act on that information.
The seriousness of the issue depends greatly on the type of meeting. A casual team catch-up about next month’s rota is one thing. A board meeting, charity trustee meeting, safeguarding discussion, HR meeting, disciplinary hearing or confidential strategy session is something entirely different. In those situations, inaccurate notes are not just an irritating administrative mistake. They can affect reputations, responsibilities, decisions and potentially people’s rights.
The answer is not to ban AI because it can be genuinely useful by saving us time like potentially reducing the burden of producing notes after lengthy discussions. The sensible approach is to treat it like any other tool: useful, efficient and requiring someone to operate it properly – this process does not need to be complicated: tell people before the AI note taker starts. explain what system is being used and what happens to the information. People should be allowed the opportunity to raise concerns, accept to receive a copy of the notes (because guess what ? quite a few systems now require you to create an account to view them and the next thing you know is that it is now sending you emails…arggghhhh!) From a personal point of view, I don’t want either the responsibility of reviewing the output, correcting mistakes and then rechecking the final version again or a hissy fit from someone who has taken umbrage at me saying no to their little AI PA.
However, I recognise that the future will almost certainly involve more AI assistants sitting quietly in meetings. The challenge is making sure they remain assistants rather than uninvited witnesses with a surprisingly creative imagination. Because the last thing any organisation needs is a meeting where nobody remembers exactly what was agreed, followed by minutes confidently explaining what happened in a conversation that existed mostly inside the AI’s imagination.
FAQs
Can someone legally record a meeting without telling everyone in the UK?
The legal position depends on the circumstances. A person recording a conversation they are part of is not automatically committing a criminal offence, but workplace and professional meetings involve additional concerns around privacy, confidentiality and data protection.
Are AI meeting notes legally binding?
No. AI-generated notes are not automatically an official record. They should be checked and agreed before being relied upon as accurate minutes.
Who is responsible if AI meeting notes are wrong?
The person or organisation using and distributing the notes remains responsible for how that information is used. Blaming the AI does not remove accountability.
Do attendees have to consent to AI note takers?
Consent is only one possible legal basis under UK GDPR, but transparency is a key requirement. People should normally be informed that AI processing is taking place and how their information will be handled.
Can AI meeting notes create employment or legal problems?
Yes, particularly if inaccurate notes wrongly attribute comments, decisions or actions to individuals, or if confidential information is mishandled.
References
Information Commissioner’s Office (ICO) guidance on artificial intelligence and data protection.
UK General Data Protection Regulation principles relating to transparency, accuracy and accountability.
Data Protection Act 2018.
ICO guidance on accuracy and records containing personal information.
