Meeting Recording Became Default. Nobody Asked.
It wasn't better tech that broke the taboo — it was a new justification.
Opening
Dear reader, here’s something one San Francisco venture capitalist has been dealing with lately. When they walk into a meeting, founders now set their phones on the table and inch them across — the way microphones cluster around an athlete right after the final whistle. A few years back, people would ask before hitting record. Now, no one asks.
This is the scene The Wall Street Journal recently reported from Silicon Valley. Just a few years ago, recording a meeting was treated with caution — a security concern. Now, not recording is what looks strange. Let me give you the conclusion up front: what broke this taboo wasn’t better recording technology. It was a new justification. And Korea is the country that completed that shift in justification first — and most quietly — in the world.
The Meeting Room Where Asking Became Rude
The scenes the Journal captured are strikingly specific. One startup founder wore a recording badge that lit up red at a work conference and got complaints that it made people uncomfortable. He said the complaints were what actually bothered him. The same person records most of his dates too, then afterward asks an AI how much he talked and how he could be more empathetic. The CEO of another AI startup says that when people in the office later realize a conversation wasn’t recorded, everyone feels a pang of regret. He added: asking permission to record kills the mood.
The venture capitalist mentioned above described the situation as society starting to resemble East Germany. There are people on the other side, too. One New York investor, believing that covert recording shouldn’t be socially acceptable, simply changed the name displayed on their Zoom account — appending a line stating they don’t consent to transcription or recording. That someone would rather bake a refusal into their display name than say it out loud every single time tells you how burdensome declining has become.
Here’s what matters: the legal situation in the US is far stricter. California and roughly 10 other states require all-party consent to record a private conversation1. That’s why Apple didn’t add a call-recording feature for 17 years after launching the iPhone in 2007. In other words, the practice is collapsing first in the country with the stricter law.
The poster child of this trend is Granola. No bot joins the meeting as a participant; it runs quietly on the user’s own computer, transcribing the conversation in the background. No icon appears on the Zoom screen, no notification sound plays. Founded in 2023, the company raised $125 million in March 2026 at a $1.5 billion valuation — 6 times the $250 million valuation it had just a year earlier. Revenue grew 250% in the most recent quarter. Existing collaboration tools like Notion, Zoom, and Teams have rolled out the same feature, but Granola keeps getting chosen because of one difference: “no bot joins the meeting as a participant — it runs quietly on the user’s own computer, transcribing the conversation.”
Korea Already Ran This Drill
Reading this piece, though, I had a slightly different reaction. For Korean readers, none of this is news. We passed through this gate a long time ago.
Korea’s Protection of Communications Secrets Act only bans recording a private conversation between other people that you’re not part of2. Flip that around: if you’re a party to the conversation, recording it without the other person’s knowledge carries no penalty. The Supreme Court has consistently ruled the same way. Conversely, secretly recording a conversation you’re not part of carries a prison term of 1 to 10 years — with no fine option at all, which shows how seriously it’s treated.
Under this legal structure, Korea has become a country where the single most common piece of evidence submitted in labor disputes is an unauthorized recording transcript. One labor attorney has described the country as having turned into a recording paradise. In 2022, a bill to penalize recording without the other party’s consent was introduced — and withdrawn under public backlash. What citizens defended wasn’t their privacy. It was their right to record.
Of course, this freedom has a line. In 2023, the Supreme Court convicted a civil servant who recorded a conversation between their boss and a visitor in the office. The defendant argued it was a public space — a civil affairs office — and that the recording was meant to expose misconduct, but the court didn’t buy it. That single step between a conversation I’m part of and a conversation between other people is exactly what separates legal from a prison sentence. And it’s precisely where AI note apps get murky.
Hardware tells the same story. Call recording has been a default feature on Galaxy phones from the start. Apple didn’t open it up in Korea and a handful of other countries until iOS 18.1 in October 2024. But here’s the detail worth noticing: when you start recording on an Apple device, it automatically plays a voice notice telling the other person the call is being recorded — even in Korea, where single-party consent makes that notice legally unnecessary.
Same act — recording — but opposite postures. Apple volunteered a disclosure it wasn’t obligated to make; AI note apps switch on as quietly as the law will allow. That’s not a difference in technology. It’s a difference in justification. Naver’s Clova Note has become an essential app for many office workers, and the Korean startup Tiro has emerged as another major player in the space.
The Moment Defense Turned Into Convenience
There’s a clear reason recording was never taboo in Korea for so long: it was a weapon. Workplace abuse (gapjil, the abuse of hierarchical power against subordinates), sexual harassment, unjust orders, unpaid wages — recording was often the only evidence the weaker party could hold onto. Defensive recording has three traits: it’s exceptional, it’s deliberate, and the moment you hit record, tension enters the relationship. That’s why people handled it carefully.
What AI note apps changed isn’t the technology’s capability — it’s this motive. Drafting minutes, organizing action items, recovering context you missed. Recording is no longer an act of self-protection; it’s become an act of doing your job well. Once the justification shifted from defense to convenience, all three traits flipped. It’s constant, it’s automatic, and switching it on requires no resolve at all. If anything, it’s turning it off that now needs an explanation.
The numbers show the shift. As of March 2026, cumulative sign-ups for Naver’s Clova Note had passed 6.6 million — up roughly 24% in a year. Naver Works Clova Note, the enterprise version launched in October 2024, has seen paying customers grow more than 40% month-over-month on average. SK Telecom’s A. Note passed 300,000 users within a week of launch. With Korea’s economically active population at roughly 29 million, Clova Note alone covers roughly one in five working people.
What really catches my attention here is something else: Granola changed what it calls itself. Since 2026, the company describes itself not as a meeting-notes app but as an enterprise context layer. In February, it added MCP integration, letting whatever’s said in a meeting flow into other AI tools like Claude or ChatGPT3. The default setting for data training is also opt-in by default — you can turn it off, but most people just leave the default as is.
Leaving it running creates its own gap. The app doesn’t know when the meeting has actually ended. Even if I briefly step out, the mic stays open, and whatever the remaining people say during that stretch is no longer a conversation I’m part of. In legal terms, it drifts much closer to a conversation between other people. Switching it on stopped requiring resolve — but so did switching it off. That’s the quietest side effect of the shift from justification-as-defense to justification-as-convenience.
Here’s the summary: what this category sells is a summary. What it accumulates is every word spoken inside an organization. Users are paying with a corpus in exchange for a summary. And more than half of that corpus isn’t my voice — it’s the voice of whoever’s sitting across from me. I can only consent on my own behalf, but what’s actually being handed over belongs to someone else.
Smart Glasses Are Standing Where Recording Stood a Year Ago
Smart glasses are following this exact path. Ray-Ban Meta and Oakley Meta officially launched in Korea on May 25, 2026. Prices start at ₩690,000, and they’re sold not in electronics stores but in department stores and optical shops. Manufacturer EssilorLuxottica says it sold more than 7 million AI glasses in 2025 alone — more than 3 times the 2 million units sold across 2023 and 2024 combined.
Friction has grown alongside it. In May 2026, 2 cases of cheating using AI glasses were caught at Korean TOEIC testing centers — the first such incidents in the history of a certified English test — and the fact that proctors had to rely on visual suspicion alone says something about the nature of this device. The BBC reported cases of men wearing these glasses secretly filming women in public and posting the footage online. Meta is preparing an update, due in July 2026, that would detect devices with a tampered recording indicator light and block filming.
And the spread is just beginning. Google is preparing Android XR glasses in partnership with Samsung Electronics and Gentle Monster, and Apple has signaled it will enter the race too. Researchers estimate that if these companies reach sales volumes close to Meta’s, as many as 100 million people could own smart glasses within a few years. It’s taken less than 2 years for the scene of phones sliding across a conference table to migrate onto a device strapped to someone’s face.
Regulation moved during the same window. On July 3, 2026, Korea’s Personal Information Protection Commission unveiled its third basic plan for personal-data protection, opening the door for lawfully collected voice and video data to be used for AI training in its raw, unpseudonymized form, subject to a risk assessment4. Detailed standards haven’t been finalized yet, so this shouldn’t be read as a settled conclusion. Still, the direction is clear: users are handing over their own voices, and the system is moving toward widening how those voices can be used.
Oz’s Lens
While building go-to-market strategy, I’ve watched adoption resistance break down more than once. It never broke because the product got better. It broke the moment a structure formed where not using it put you at a disadvantage. AI note apps have arrived at exactly that spot: a situation where only I lack meeting minutes, only I can’t search past meetings. The instant that asymmetry appears, individual privacy sensitivity loses its grip. Norms don’t collapse because people turn bad — they collapse once the structure makes whoever keeps the norm the one who loses out.
So does this mean we trust each other more, or less? I don’t think it’s either. It’s closer to trust becoming unnecessary. Trust is the act of skipping verification — and once verification becomes free, there’s no reason to skip it anymore.
The trouble is that a record is not a full substitute. A record only captures what got said — but what’s genuinely valuable in an organization is often what didn’t get said. Half-baked ideas, real dissent, early warnings of bad news. The moment these get recorded, they carry a cost, so they flee outside the record. It’s more accurate to say trust hasn’t shrunk — it’s been relocated. From the conference room to the hallway. The more perfect meeting minutes become, the more the paradox emerges that the decisions that matter get made outside the minutes.
That said, I’ll resist the nostalgia that says people used to trust each other. The very fact that Korea’s recording culture began as a defense against workplace abuse and sexual harassment refutes that nostalgia. People didn’t trust each other back then either. The weaker party simply had no evidence.
Closing
Three lines to close on. What broke the taboo around meeting recording wasn’t better technology — it was the justification of productivity. Korea had already completed that shift, thanks to legal single-party consent and a call-recording culture. And a substantial share of the data handed over in exchange for that convenience isn’t mine — it belongs to the person sitting across from me.
Here’s one thing worth doing this week: check whether the data-training setting is switched on in whatever note app you use. For most, the default is on.
Which side are you on, dear reader? If you’re someone who turns on an AI note app during meetings, do you tell the other person first? And if you’ve been on the other side — realizing someone else had it running — how did that feel? I’m especially curious about any moment you hesitated over whether to hit record. Share it in the comments, and once enough stories come in, I’ll pull together a read on the real temperature of Korean meeting rooms in the next issue.
💬 Tell me in the comments whether you let the other person know before turning on an AI note app. I’ll fold your answers into the next issue. 📨 If you have a colleague who sits in a lot of meetings, pass this along to them.
References & Further Reading
Primary sources
- The Wall Street Journal, “This Conversation Is Being Recorded. They All Are.”, 2026. Link ··· This is where today’s piece starts. The scene of someone recording even a date to get AI feedback afterward is the standout moment.
- Seoul Economic TV, “Clova Note’s Paid Customers Grow 40%: AI Meeting-Notes Revenue Kicks Into Gear,” 2026.3.31. Link ··· This is the source for Clova Note’s 6.6 million cumulative sign-ups and its enterprise growth rate. It’s a good way to gauge the scale of the Korean market.
- TechCrunch, “Granola raises $125M, hits $1.5B valuation as it expands from meeting notetaker to enterprise AI app”, 2026.3.25. Link ··· This article lays out the shift from meeting-notes app to context layer.
- Personal Information Protection Commission, “The Third Basic Plan for Personal Information Protection to Promote Trust-Based AI Innovation (2027–2029),” 2026.7.3. Link ··· This contains the direction on using raw voice and video data for AI training. Detailed implementation standards haven’t been released yet, so read it with that caveat.
Background
- Kwangsun Lee, “An Employee Who Records Every Conversation and Call — Can the Company Discipline Them?,” Hankyung CHO Insight, 2023.11. Link ··· This lays out, from a practitioner’s perspective, why Korea became a recording paradise and where the legal boundary of single-party recording sits.
- Edaily, “iPhone Call Recording Arrives This Month — Supported on iPhone XS and Later,” 2024.10. Link ··· This confirms the background on why Apple added a recording notice in Korea despite having no legal obligation to do so.
- Platum, “Smart Glasses Finally Become Invisible,” 2026.7. Link ··· Covers the Korean launch and the TOEIC cheating cases in one place, giving a full view of where glasses are headed.
Past issues worth reading alongside this one
- The People Who Switch Off Their 20-Watt Brains ··· This covered what changes when we hand judgment over to AI. If today’s piece is about outsourcing memory, that one is about outsourcing thought.
- Why Does the Rooftop-Room Developer Deliver Food Every Night ··· A look, from the labor side, at how the justification of convenience reshapes an individual’s circumstances.
📝 Glossary
Footnotes
-
All-party consent: The principle that recording a private conversation requires the consent of every participant. Standards vary by US state — California and roughly 10 others require this approach, while the rest allow recording with just one party’s consent. ↩
-
Protection of Communications Secrets Act, Article 3: The provision banning the recording or interception of a private conversation between other people that you’re not part of. The key phrase is “between other people” — a conversation you participate in doesn’t fall under this rule. That said, whether any specific case is unlawful depends on the circumstances, so real disputes call for expert consultation. ↩
-
MCP (Model Context Protocol): A standardized specification that lets AI tools connect to external data or other services. Once a meeting-notes app adopts this, a channel opens for accumulated conversation records to flow into other AI services. ↩
-
Pseudonymization: The process of removing or altering identifying information — like names or contact details — so that a person can’t be identified without additional information. Until now, this has generally been a required step before personal data could be used for AI training. ↩


Your take shapes the next issue
Reply with your experience or perspective — the best responses feed into future issues.
Sign in to commentAny registered reader can comment — it takes 10 seconds.