OPSEC for AI tools
Check the information in a request, not just the topic of its answer.
In this lesson · 02 of 7
All lessons in this track →Saved in this browser. View your learning.
Start with the idea
Check the information in a request, not just the topic of its answer.
A public question can arrive with sensitive surrounding details. Do not transfer those details without permission. Provider settings about training and retention are different from organizational approval.
Data path: The route information takes through inputs, tools, logs and outputs.
See the example and check your understanding
A made-up teaching example
The training log combines a public office-hours question with internal-looking host and shift details. Practice the public question separately with invented facts, rather than transferring the whole log.
Ask yourself
Does a public answer make every attached log field safe to upload?
Show the explanation
No. Check the whole input and its ownership. “No training” does not mean “no retention” or permission to share.
Start with these checks
- Review all included fields.
- Hold unresolved ownership or permission.
- Keep invented training alternatives available.
Use the different practice case below and check every requirement in its full instructions.
The video explains the main idea. Use the sources, instructions and checks below to complete the practice.
Captions appear here during playback.
OPSEC for AI tools
Check the information in a request, not just the topic of its answer.
Video playback is unavailable. Read the descriptive transcript below, go directly to practice, or open the video file.
Source-based teaching example · Fictional data · Reviewed 2026-10-03. Microsoft Mark · locally synthesized. On a small screen, use fullscreen for diagram details or read the transcript.
Read the descriptive transcript
0:00 · What you will learn
Check the information in a request, not just the topic of its answer.
What appears on screen
Check the information in a request, not just the topic of its answer.
0:07 · The idea
A public question can arrive with sensitive surrounding details. Do not transfer those details without permission. Provider settings about training and retention are different from organizational approval.
What appears on screen
A public question can arrive with sensitive surrounding details. Do not transfer those details without permission. Provider settings about training and retention are different from organizational approval.
0:22 · One useful term
Data path. The route information takes through inputs, tools, logs and outputs.
What appears on screen
The route information takes through inputs, tools, logs and outputs.
0:31 · A worked example
The training log combines a public office-hours question with internal-looking host and shift details. Practice the public question separately with invented facts, rather than transferring the whole log.
What appears on screen
The training log combines a public office-hours question with internal-looking host and shift details. Practice the public question separately with invented facts, rather than transferring the whole log.
0:47 · Your turn to think
Does a public answer make every attached log field safe to upload? Pause and decide before continuing.
What appears on screen
Does a public answer make every attached log field safe to upload? Pause the video if you need more time.
1:04 · Why this answer works
No. Check the whole input and its ownership. “No training” does not mean “no retention” or permission to share.
What appears on screen
No. Check the whole input and its ownership. “No training” does not mean “no retention” or permission to share.
1:15 · Check 1
Review all included fields.
What appears on screen
Review all included fields.
1:20 · Check 2
Hold unresolved ownership or permission.
What appears on screen
Hold unresolved ownership or permission.
1:25 · Check 3
Keep invented training alternatives available.
What appears on screen
Keep invented training alternatives available.
1:31 · Now try a different case
Open the lesson and try its different practice case. Use every check in the full task. Watching alone does not mark the lesson complete.
What appears on screen
Open the lesson Try the new case Check and revise The full task and its requirements are in the linked lesson.
Watching explains the example. Complete the lesson practice to check what you can do. Explore all visual lessons →
Decide before disclosure. Build a judgment you can explain and a result another reader can inspect.
Prerequisites: know the difference between generated text and source evidence. For configuration vocabulary, use the AI workflow reference. This is public self-study, not official government training, accreditation, a clearance or a professional credential.
Observable outcomes: Apply a pre-disclosure decision to six synthetic prompt cards; distinguish training, retention, aggregation and authorized use.
Part 1 Decide before disclosure
A consumer chat service is not literally a public forum. Provider training defaults, retention, administrator access, contractual protections and product settings vary. None of those facts independently authorizes protected work. Before transfer, establish the information owner, handling requirements, exact service and tenant, user role, purpose, destinations and organizational approval. An unmarked file is an unresolved question, not public-release evidence.
For classified, CUI, export-controlled, personal or employer-confidential information, use the responsible security/privacy/export-control route and the exact approved system. A FedRAMP label is not blanket permission for CUI or classified material. Consider aggregation: several public-looking facts can expose a protected relationship. For suspected exposure, stop the affected activity and follow the designated incident process; do not export, delete or investigate sensitive history as a classroom exercise.
Part 2 Synthetic worked example
Synthetic log: service=ORION, fictional internal host=node-17, schedule=night-shift, question=public office hours. The public question can be practiced separately using invented office hours. The log should not be transferred merely because its answer concerns public information. A training opt-out changes one provider use; it does not settle logs, access or agency permission.
Guided practice: write the tempting conclusion first, then underline the fact that does not support it. Rewrite the conclusion with that limitation visible. Identify one additional observation that would change the decision.
Next step: carry your artifact into deployment boundaries in lesson 03; retain assumptions and unanswered questions.
Source: NARA CUI program, consulted 2026-10-02. Agency implementation and the information owner govern real handling decisions.
Provider example: OpenAI data-use documentation, consulted 2026-10-02, distinguishes consumer settings and business-service defaults. Check the exact product and current agreement; this example is not authorization.