Surveillance isn't about what gets recorded — it's about new capture you didn't know about. Google has been recording every keystroke of every student essay for over a decade. Rumi doesn't add a camera; it just turns on the lights in a room that was already being filmed.

Most people don’t know this: Every Google Doc you’ve written since 2010 contains a microsecond-accurate record of every key you ever pressed. (source)

Every word you deleted, every awkward phrasing you backspaced over, and every 2 a.m. struggle with a thesis statement is preserved in a permanent, append-only ledger. Since Microsoft Word’s "Cloud-First" shift in late 2025, the same is now true for Word; documents are streamed to the cloud from the very first keystroke, often before you’ve even given the file a name.

This isn’t a leak or a conspiracy. It’s the standard infrastructure of modern writing—and it affects almost everyone on Earth.

The Word Processor Duopoly

As of 2026, Google Workspace and Microsoft 365 control more than 95% of the office productivity market. In education, Google alone is the "de facto" platform for 170 million students and teachers. If you wrote it on a computer in the last fifteen years, you weren't just using a tool; you were contributing to a ledger you didn't know existed.

We bring this up because the most common concern we hear about Rumi is that capturing the writing process feels "invasive." But that concern rests on a misunderstanding of what surveillance actually is.

Surveillance vs. Insight

When we think of surveillance, we usually imagine a camera being turned on. We notice the "new" thing. Because Rumi explicitly shows you the writing process, it feels like "new" tracking.

But true surveillance is often the air pressure—the constant, invisible recording we’ve stopped noticing.

  • Passive Recording (Surveillance): This is what Google and Microsoft do. It happens behind a one-way mirror. The data is collected silently, stored in a corporate vault, and used for AI training or analytics. You have no visibility into what they know about your struggle to write.
  • Active Disclosure (Insight): This is what Rumi does. It’s a two-way mirror. We turn the "projector" around so the student sees exactly what the platform sees.
Surveillance isn't the act of recording; it is the act of keeping that recording a secret from the person being recorded.

Your Writing Is Already Being Recorded

As programmer James Somers demonstrated, a Google Doc is a living ledger. Every character has a persistent unique ID. If you cut a sentence from page 10 and paste it on page 1, that character’s ID—and its original timestamp—remains.

If you’ve ever used the Draftback extension to play back a Doc like a movie, you’ve seen this "shadow data" in action. Draftback didn't invent that movie; it just hit "Play" on a film Google had already recorded. With the 2025 shift to "Cloud-First" defaults, Microsoft has gone a step further, making every draft "Copilot-ready" from the first keystroke.

The data layer

Who's already recording your keystrokes

Three platforms, the same underlying telemetry. The difference is who gets to see it — and what it's used for.


Dimension
Platform
Google Docs
Platform
Microsoft Word
Rumi
Platform
Rumi
Captures keystrokes
Yes — every mutation, microsecond-accurate, since ~May 2010
Yes — continuous AutoSave to OneDrive; cloud-first from first keystroke (late 2025)
Yes — keystroke-level telemetry tied to a specific assignment
Disclosure to writer
Shadow data
Behind an undocumented endpoint; most users never see it
Documented
AutoSave toggle visible; Version History in support docs
Disclosed up front
Stated in ToS and in the product before the first keystroke
Who can see the process
Anyone with Edit permission — via Draftback or similar (500K+ users)
Microsoft, plus version-history viewers; surfaced for recovery, not understanding
The student first; the instructor for assignments already being submitted to them
What it's used for
Collaboration infra, version recovery, AI training, product analytics
Recovery, version history, making documents "Copilot-ready" by default
Process visibility for the writer; academic-integrity context for the instructor
Used to train foundation models
Per platform terms
Copilot-ready by default
No
Google Docs
Captures keystrokes
Yes — every mutation, microsecond-accurate, since ~May 2010
Disclosure to writer
Shadow data
Behind an undocumented endpoint; most users never see it
Who can see the process
Anyone with Edit permission — via Draftback or similar (500K+ users)
What it's used for
Collaboration infra, version recovery, AI training, product analytics
Used to train foundation models
Per platform terms
Microsoft Word
Captures keystrokes
Yes — continuous AutoSave to OneDrive; cloud-first from first keystroke (late 2025)
Disclosure to writer
Documented
AutoSave toggle visible; Version History in support docs
Who can see the process
Microsoft, plus version-history viewers; surfaced for recovery, not understanding
What it's used for
Recovery, version history, making documents "Copilot-ready" by default
Used to train foundation models
Copilot-ready by default
Rumi
Rumi
Captures keystrokes
Yes — keystroke-level telemetry tied to a specific assignment
Disclosure to writer
Disclosed up front
Stated in ToS and in the product before the first keystroke
Who can see the process
The student first; the instructor for assignments already being submitted to them
What it's used for
Process visibility for the writer; academic-integrity context for the instructor
Used to train foundation models
No

What This Means for Classrooms

The real surveillance problem in education today is presumed guilt. AI detectors return a percentage and a verdict: "This looks like AI." They force students to defend themselves against a black-box algorithm. That is a "spy camera" model.

Process visibility is a "mirror" model. It doesn't watch the person; it respects the evolution of the text. Genuine intellectual work leaves a trace that AI cannot retroactively manufacture: the rhythmic pauses, the dead-end paragraphs, the arguments that shift direction as the writer learns.

When that trace is visible, the student stops needing to "prove a negative" to an AI detector. Instead, they have a verified receipt of their own thinking.

The data exists either way. The only question is who it serves.

Giving Students the Projector

The writing process is the only honest artifact of how thinking happens. For decades, Google and Microsoft have been keeping the film of your work in a vault. Rumi just gives the student the projector.

You aren’t being "newly watched." You’re finally being shown what was already there—and for the first time, it belongs to you.

Learn how Rumi supports AI Literacy and Academic Integrity