I was just in a meeting where someone used a thing called Fathom to get an 'AI' summary of the meeting. Aside from some understandable typos arising from not understanding terms of art and replacing them with common English words, one of the key points that it concluded was that A was faster than B. It reached this conclusion because it missed one of the digits in the time for A. This completely inverted the key takeaway from one important section of the meeting.

Do not use plausible-nonsense generators for anything important.

@david_chisnall
Another risk is that these LLMs can be easily influenced to give more weight to an opinion rather than others.

https://www.schneier.com/blog/archives/2025/11/ai-summarization-optimization.html

AI Summarization Optimization - Schneier on Security

These days, the most important meeting attendee isn’t a person: It’s the AI notetaker. This system assigns action items and determines the importance of what is said. If it becomes necessary to revisit the facts of the meeting, its summary is treated as impartial evidence. But clever meeting attendees can manipulate this system’s record by speaking more to what the underlying AI weights for summarization and importance than to their colleagues. As a result, you can expect some meeting attendees to use language more likely to be captured in summaries, timing their interventions strategically, repeating key points, and employing formulaic phrasing that AI models are more likely to pick up on. Welcome to the world of AI summarization optimization (AISO)...

Schneier on Security