With AI keeping various board meeting records, who owns the truth?

For decades, board-meeting minutes were a study in understatement. A diligent secretary would jot down the essentials: agenda followed, motions made, votes cast and resolutions approved. No frills, no flourishes. Minutes were often praised for being so dull, they could cure insomnia. Fast-forward to today, and suddenly technology has barged into the boardroom, not just to help take notes, but to interpret, condense and sometimes even decide what counts as ‘important.’

Sounds efficient? Perhaps. But as with all things AI, the devil isn’t just in the details—it’s in the transcripts, summaries and the alternative versions floating around in cyberspace. And directors are beginning to realize that minute-taking could become its most litigious landmine.

Here’s a story that has achieved the status of an urban-legend. A director logged into an online board meeting but wasn’t actually present. He had his AI assistant attend in his place, dutifully summarizing the proceedings for him to skim later. Efficient? Yes. Fiduciary? Questionable. Legal? Problematic.

Can a director truly be said to have ‘attended’ if his AI assistant took the notes while he took a nap? What about fiduciary duty—can responsibility be delegated to a bot? Can you trust an AI tool, operating on remote servers, to safeguard sensitive strategy discussions? Litigators could have a field day.

Even when directors are actually present and engaged, the nightmare need not end. Consider a common transcription widget—an app quietly humming away on a laptop or phone, producing a verbatim record of everything said. While meeting platforms usually give its moderator control over recordings, nothing prevents a determined director from running a parallel transcript.

Now picture the legal discovery process five years down the line. The official board-approved minutes state one version of events, while a director’s personal AI transcript stored in the cloud tells another. Which one prevails? Suddenly, the carefully sanitized official record is in conflict with a rogue version full of side comments, half-formed ideas and jokes that should never have been aired. Guess which one lawyers will demand.

Whose summary is the ‘truth’? AI summary tools are multiplying faster than mushrooms in the monsoon season. Each promises to distil meetings into crisp action points. But when one AI insists that the key takeaway was a budget reallocation, another claims the real highlight was an acquisition plan and a third focuses on a throwaway remark about “maybe moving to Indonesia.” Which one is right?

Directors could find themselves in an absurd situation: four different summaries of the same board meeting, each with its own version of ‘truth.’ It’s the Rashomon effect for governance. In Akira Kurosawa’s classic film, the same event is told from four conflicting perspectives. In today’s boardrooms, AI is unintentionally staging its own version. Unlike Kurosawa’s Rashomon, where ambiguity was philosophical, here it’s potentially actionable. A stray AI-generated line can change the meaning of a discussion, or worse, suggest that the board ignored a key risk.

Then comes the thorniest issue of all: confidentiality and privilege. Lawyers already walk a tightrope when it comes to what goes into minutes. Too much detail and a sensitive legal strategy could be exposed. Too little, and it looks like the board glossed over critical issues. Imagine an AI tool diligently transcribing every lawyer-client exchange, summarizing it and uploading it to the cloud. Privilege, meet the shredder. What a nightmare!

So why are directors tempted by these tools? Because the lure is real. Note-taking is tedious. Summaries are useful. And AI can do in seconds what a human might take hours to polish. For a time-strapped director juggling ten committees, three companies and one golf handicap, automated summaries are irresistible.

There’s also the optics of efficiency. Board packs are already bloated, sometimes running into hundreds of pages. Directors are desperate for tools that cut the clutter. AI, with its bullet-point brilliance, seems like salvation. But as any magician will tell you, the trick looks clean only because you’re not watching closely.

Where does this leave boards? Not with the comfort of dos and don’t, because AI is evolving too fast for hard-and-fast prescriptions. But a few thought starters are clear.

Transparency matters: Boards must discuss whether AI note-taking is okay, in what forms and with what safeguards. Consistency counts: If official minutes say one thing and rogue transcripts say another, the credibility of the board is at risk. Alignment is a must for good governance.

Guard privilege like treasure: AI tools are only as safe as their servers. A single slip could expose privileged legal advice. Embrace the human factor: Ultimately, minutes are not just about record-keeping; they are about accountability. No AI tool can bear fiduciary responsibility.

Recall Mark Twain: “It’s not what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so.”

The authors are, respectively, co-founder of the Medici Institute for Innovation.X: @MuneerMuh; and a global board advisor, coach and publisher.