Google's Drive update lets Gemini label confidential files from plain-language rules
As Gemini does more tasks on its own, Google released a Drive tool that tags sensitive files automatically, with no model training by the admin.
The short version
- Google announced two Gemini updates on November 17, 2026: Tasks in Gemini going free, and Drive AI classification
- Admins write plain-language rules and Gemini labels files by sensitivity
- The older method needed at least 100 hand-sorted training files per label
- AI labels never override DLP rules or labels a person set or approved
- Rollout began November 17, 2026 and should finish on December 15, 2026
Why it matters
The more freely an AI assistant can act across your files and apps, the more it matters that sensitive files are clearly marked. This update makes that marking much cheaper for companies. If your employer has put off labeling files, this removes one of the main excuses.
Two updates, one theme
As kn.itmedia.co.jp reported, Google announced two Gemini updates on November 17, 2026. One widens autonomous task-doing for everyday users. The other helps companies keep confidential data under control.
On the consumer side, Tasks in Gemini, which runs on a model called Gemini 4 Hydrogen, became free in 6 countries including Japan and the US. It handles multi-step jobs in the background, such as a morning briefing built from your email and calendar, and notifies you when it finishes.
What Drive's new labeling does
The business-side update is Gemini-powered AI classification in Google Drive, now generally available. Gemini reads a file's content and attaches a data classification label, such as a confidentiality level.
Before, an admin had to sort at least 100 example files by hand for each label and train a machine learning model. Now the admin simply writes instructions in everyday language: what kind of document gets which label, plus any exceptions.
Guardrails around the guardrail
The AI does not overwrite DLP (data loss prevention) rules, or labels that a person set or approved by hand. Admins can test results on sample text in the admin console before going live. Afterward they can watch label counts and user corrections almost in real time.
There are limits. Files need readable text, which includes text pulled out by OCR. Files under client-side encryption, plus audio and video, are excluded. Admins can create up to 5 models, counting older custom ones. Google says file contents and instructions stay inside the organization and are not used to train its global models.
Who gets it and when
It is available on Enterprise Plus, Google AI Pro for Education and Frontline Plus. Rollout started on November 17, 2026 and is scheduled to finish on December 15, 2026.
What to watch
- How accurate the automatic labels prove in real company use, and how often people correct them
- Whether the external-connection features of Tasks in Gemini, now US-only, reach other countries
Try it
If you use Google Workspace at work, ask your IT team whether sensitive files carry labels, and review which files and apps Gemini can reach for you.
Sources and links
First reported by kn.itmedia.co.jp. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.