August 27, 2026
I Left the Office for Three Months. AI Didn’t.
Three months away from work. I came back to find an AI assistant in every tool - and a security officer's view of what nobody risk-assessed.
By Oleksii Medvedev
3 min read
I Left the Office for Three Months. AI Didn't. (*)
Bill Gates published an essay this week — "The turbulent AI era is here. The choices we make now are critical." arguing that the transition into the AI era will be one of the most turbulent periods in human history, and that the world is not adequately prepared for it. He writes about it at the scale of nations, labor markets, and international institutions.
I can offer a much smaller data point.
Earlier this year, I was away from work for about three months. When I returned, I got to experience something almost nobody gets to experience deliberately: seeing several months of AI adoption all at once, compressed into a single Monday morning.
Gates argues that we need better preparation for the AI era. From ground level, I saw how difficult it is for governance to keep pace with the speed of adoption.
The workplace I came back to
Before I left, AI at work was a thing you went to — a chat window, a specific tool, a deliberate act.
When I came back, it was a thing that came to you.
Every application seemed to have sprouted an assistant. Email offered to summarize threads I had already read. Meeting notes could write themselves. Document editors, ticketing systems, browsers, even collaborative tools were increasingly offering to summarize, explain, draft, and "help."
As a user, I'll admit it: some of it is genuinely useful. Catching up on several months of email is a very different task when a machine can help triage it.
As an information security professional, I had a different reaction.
Every one of those helpful little buttons is a data flow.
And the interesting change isn't simply that there are more AI tools. It's that AI capabilities are becoming embedded in tools people already use, often with very little deliberate decision-making required from the user.
What Gates calls misuse, I call Tuesday
In his essay, Gates identifies AI-enabled fraud and cyberattacks among the major risks of this transition.
From where I sit, one of the more immediate challenges isn't the attacker with an AI — it's the colleague with one. Not malicious. Helpful.
Consider what happens when an employee pastes sensitive business information into an assistant to "just get a quick summary."
Or when an AI notetaker joins a meeting where confidential information is discussed.
Or when a new integration becomes available because a software vendor enabled it by default.
None of this is new in kind. Shadow IT is as old as the USB stick.
What's new is the speed and the default-on nature of it.
Shadow IT used to require effort: you had to go find the unauthorized tool. Shadow AI increasingly requires effort to avoid.
The assistant is already inside the software you licensed. Declining to use it can become the deliberate act.
Meanwhile, the governance layer moves at governance speed.
Vendor assessments, DPIAs, sub-processor reviews, records of processing, security reviews — the paperwork of trust can take weeks or months.
The tools can change in days.
That's the uncomfortable part.
Gates argues that the choices made during the early years of the AI era will have consequences far beyond the technology itself. It's easy to read that as an abstract policy debate.
But the same argument applies one floor down, inside every company.
The choices your organization makes about AI this quarter (which assistants are sanctioned, what data may touch them, who reviews the defaults vendors switch on, whether AI processing is properly reflected in privacy documentation) will influence its risk posture for years.
Retrofitting governance onto tools that people already love is far harder than shaping adoption while it's happening.
The challenge is not to stop AI adoption. That train has already left the station.
The challenge is to make sure that security, privacy, compliance, and responsible use are travelling on the same train.
We don't get to wait for the international institutions Gates is calling for.
The AI era didn't ask the governance calendar for a slot.
It arrived while I was away, embedded itself into everyday workflows, and greeted me on my return with an offer to summarize what I'd missed.
I declined.
Some things you need to read yourself.
Bill Gates, "The turbulent AI era is here. The choices we make now are critical."
* — The views expressed here are my own and reflect general observations about AI adoption, information security, privacy, and governance across the industry. They are not a description or assessment of any specific employer, organization, system, or security environment.