The Three Seconds Before You Paste

Originally written in Traditional Chinese: 公司在擋的不是 AI：是「貼上」之前，沒有人想過的那三秒

As of Monday, staff computers at Sweden's Government Offices can no longer open ChatGPT or Claude in the browser. Many will read that as fear of AI. I see something else.

The Government Offices, which include the Prime Minister's Office and the ministries, began blocking the browser versions of ChatGPT and Claude on October 26. Staff are directed to Microsoft Copilot, which the Government Offices says it reviewed and then tightened the service's security. The reason given is information security: according to reports, the director of digitalization told staff that using AI services without a formal agreement can affect it. This is not a ban on AI; staff can reportedly still reach the same companies' models through an internal platform. It is a decision about which door stays open.

Last week, a hospital in Baltimore began mailing letters to 3,182 patients. On September 21, an employee at Johns Hopkins Bayview Medical Center uploaded a spreadsheet of patient information to a public AI tool the hospital had not approved. It may have included names, admission and treatment dates, medical record numbers, and medications. It was caught the same day, and the hospital asked the AI company to delete the data. The letters still had to go out.

The hospital called it a mistake. The employee probably meant no harm. That is exactly the point.

My take: organizations are not blocking AI. They are blocking the three seconds before "paste" that nobody thought about.

We usually frame AI risk as wrong answers. The everyday risk is in what you paste. That box often holds more than you: a client's name, a colleague's review, a student's grades, a patient's medication. None of them know they were pasted in.

In my course I teach seven judgment questions for deciding what to hand to AI. The one most often skipped: "If the people involved knew AI was used, would they agree?" It is not a question about you. It is about the people who are not in the room.

To be clear, I am not saying stop using AI at work, and I am not ranking which AI is safer. Sweden did not say AI is off limits either. What it blocked was the route without a formal agreement.

Try a pre-paste check. Before pasting a file, a thread, or a table, pause three seconds and ask: Does this contain anyone else's name, number, or ID? Is this tool approved by my organization? Would the people in it agree? If any answer is unclear, don't paste. Then replace names with A, B, C and delete ID columns. AI does not need to know who someone is to tidy a spreadsheet.

Educators: grade sheets are the easiest thing to paste without thinking; ask which tools your school approved. Parents: tell your child, "What you paste into AI is sometimes someone else's secret."

Sweden's closed door is not fear of AI. It is practice: before you hit paste, ask who else is in there.

Get the Human-First practice card

Four steps, ten minutes, with copy-ready AI prompts. The card and my weekly letter are currently written in Traditional Chinese.

👉 Subscribe free to 碩學苑 (Traditional Chinese)

Follow in English

I also post in English on LinkedIn.

👉 Follow Haishuo Lee on LinkedIn

References