How to Set a Personal AI Policy Across Every Note App

The annoying part of setting a personal AI policy for note taking and productivity is that the policy does not live in one place. It lives in Notion’s workspace settings, Evernote’s AI feature switches, Apple’s system-level Intelligence controls, Goodnotes’ regional and feedback settings, Notability’s Learn permissions, and whatever Obsidian or Logseq plugin you installed three months ago and forgot about.

So the setup has to start with configuration, not beliefs. Pick one risk tier, apply it across every note app you use, and write down the date you verified each vendor claim. If you want the broader product landscape first, FlowDesk’s profile of what the AI boom changed in note-taking apps is the companion piece. This guide is the settings pass.

Pick one tier before opening settings

These three tiers are not vendor product names. They are a policy layer you impose on top of messy app settings. The point is to stop treating “AI” as one moral yes-or-no question when your apps expose it as a pile of unrelated defaults, permissions, retention claims, feedback switches, and plugin choices.

The middle tier is the one that usually breaks in real life. It requires more than saying “I’ll be careful.” It needs boring capture rules: confidential meeting notes go in a non-AI workspace, student accommodations never go through lecture-summary tools, and quick notes that might contain private material do not get dumped into whichever app has the prettiest assistant button this week.

Per-app AI settings map, last verified Aug. 25, 2026

Every training, retention, and processing statement below is a vendor-stated claim unless marked otherwise. That is not the same as an independent audit. It is still useful, because settings work depends on knowing what the vendor currently says, where the switch lives, and what you should re-check before changing your workflow.

Where the map needs extra caution

Evernote: enabled by default is the setting that matters

Evernote is the row I would not skim. Its current FAQ says the v11 AI features — AI Assistant, Semantic Search, and AI Meeting Notes — are enabled by default and individually disable-able. That combination is privacy-relevant because “I never turned on AI” is not a reliable memory if the feature shipped already on. The useful part is that Evernote gives direct paths: Settings > Preferences > AI features on desktop and web, and Settings > AI features on mobile. The same FAQ states that Evernote does not train on user content and that AI Assistant data sent to OpenAI is deleted after the response; treat both as vendor claims last checked against a page updated Jan. 15, 2026. [1]

For AI-ok, that may be enough. For AI-with-redaction, the trap is meetings. A harmless task notebook and a client meeting transcript do not deserve the same setting just because they live under the same elephant logo. If Evernote is where meeting notes land automatically, disable AI Meeting Notes unless that capture path is known to be safe for your tier.

Notion: one settings page, several different questions

Notion’s privacy posture cannot be reduced to “AI on” or “AI off.” The relevant controls checked here live at Settings > Notion AI and include AI web search plus Share data to improve Notion AI. Notion states that it does not train on customer data by default, that LLM providers retain non-Enterprise data for 30 days or fewer, that Enterprise has zero retention, and that embeddings are deleted within 60 days after page or workspace deletion. Those are different claims about different parts of the system, not one blanket privacy sentence. [2][3]

That distinction matters if Notion is your shared project system. A team wiki page, a draft launch plan, and a personal habit tracker may sit in the same workspace but deserve different handling. If your tier is AI-with-redaction, the safer operational move is not to rely on memory when someone uses AI; keep sensitive pages in a workspace or section where people know AI use is restricted. For broader Notion fit, AI Agents, and pricing context, FlowDesk’s Notion Review 2026 is the better place to zoom out.

Apple Notes: control Apple Intelligence, not just Notes

Apple’s documentation frames Apple Intelligence as on-device first, with Private Cloud Compute for more complex requests, and says Private Cloud Compute does not store data. The controls checked here are system-level: the Apple Intelligence Report appears under Settings > Privacy & Security, while Screen Time > Content & Privacy Restrictions > Intelligence & Siri can block Writing Tools and third-party AI extensions. [4][5]

The practical consequence is simple: if you use Apple Notes as quick capture, your policy has to cover the whole Apple Intelligence layer. For AI-with-redaction, select and rewrite only text you have already reviewed. For AI-off, use Screen Time restrictions rather than hunting for a Notes-only switch that was not found in the checked Apple documentation.

One caution: Basil AI published a 2026 analysis estimating that shorter Writing Tools requests may stay on device while longer passages may escalate to cloud processing, and it argued that users cannot force an on-device-only mode. That is a competing local-transcription vendor’s testing and interpretation, not an Apple-documented threshold, so it should be read as a caution signal rather than a control spec. [13]

Goodnotes and Notability: students need note-level habits because class archives get messy

Handwritten study apps are easy to underestimate because the notes look casual. They are often not casual. A semester archive can contain graded feedback, disability accommodations, patient scenarios, tutoring notes, names, schedules, and photos of whiteboards that were never meant to become AI-visible material.

Goodnotes says its AI runs on an external LLM hosted on Amazon Bedrock under AWS privacy rules, that notebook data is not used for training unless voluntarily submitted through Feature Feedback, that opted-in conversation data is retained up to 30 days, and that Goodnotes AI is not intended for protected health information under HIPAA. It also says availability is region-restricted, including unavailability in China, Hong Kong, Macau, Russia, and other regions. [6][7]

Notability’s Learn documentation is narrower. It says Learn does not train AI systems or LLMs on user data, never accesses notes without explicit permission, and can be managed at Settings > AI Customization. The page was updated June 25, 2026. [8]

For AI-with-redaction, the usable rule is not “never summarize class notes.” It is: summarize the chemistry lecture, not the page with a professor’s private feedback, a classmate’s details, or clinical material. If your notebooks are already a mixed pile, clean the folder structure before trusting yourself to make the right decision from a share sheet.

Obsidian and Logseq: plugins are not defaults, but they can still leak your policy

Obsidian and Logseq sit in a different category from hosted AI-first note systems. The privacy question is less “what did the vendor enable by default?” and more “what did you install, what endpoint did you connect, and which blocks did you select?” That gives careful users more control, but it also creates a false sense of safety if a cloud API key is quietly sitting inside a plugin settings pane.

Smart Connections for Obsidian describes local on-device embeddings, no API key, and private/offline behavior by default; its plugin page also lists about 1.1 million downloads. That is a useful option for people who want AI-adjacent search without immediately routing a vault to a cloud model, but it is still a plugin choice that needs version-by-version review. [9]

For local LLM use, the key detail is the endpoint. An XDA walkthrough points Obsidian Copilot at LM Studio’s local server URL, http://localhost:1234/v1, so note content is not being sent to a cloud provider in that setup. Asian Efficiency’s setup notes give the other half of the reality check: AI plugin setup can be quick, cloud API use may cost roughly $3–10 per month for moderate use, and local Ollama models can be free but need at least 8GB of RAM, with 16GB recommended, and may produce lower-quality output. [10][11]

Logseq needs the most careful wording. The source checked here is a May 2023 forum thread saying AI features were disabled by default, individually toggleable, and sent only explicitly selected blocks or pages after review. That is useful historical evidence, not a current vendor-help-center guarantee. If Logseq is part of your AI-off or AI-with-redaction setup, verify the current build and plugin behavior before assuming the old selected-block model still describes your system. [12]

Apply the policy as a settings note

Make one small note called something boring like “AI settings — personal policy.” Put it somewhere you will actually find it when a new assistant button appears. Then record the tier, the apps, and the verification date. This does not need to become a compliance manual.

1. Choose AI-ok, AI-with-redaction, or AI-off for your whole note system. If one app needs an exception, write the exception down instead of keeping it in your head.
2. Open each app’s AI settings page and match the toggles to the table above.
3. For plugin-based apps, list the plugin name, model provider, endpoint, and whether any API key is connected.
4. For AI-with-redaction, create the capture rule before using the AI feature: which notebooks, folders, vaults, pages, or blocks are allowed?
5. Add “last verified: Aug. 25, 2026” next to each app today, then update that date only after re-checking the vendor page or current app settings.

The re-check is not optional housekeeping. Evernote can change defaults, Notion can change workspace controls, Apple can change system restrictions, and plugin authors can change model routing. Before you turn on a new summarizer, connect a new API key, or let a study app process a semester’s worth of handwriting, compare the current settings page against your policy note.

A personal AI policy is not what you believe about AI. It is what Notion, Evernote, Apple Notes, Goodnotes, Notability, Obsidian, and Logseq are actually configured to do today.

References

1. Evernote's AI Features FAQ — Evernote Help & Learning, Jan. 15, 2026
2. Notion AI security & privacy practices — Notion
3. Notion AI FAQ — Notion
4. Apple Intelligence and privacy on iPhone — Apple Support
5. Block access to Apple Intelligence features in Screen Time — Apple Support
6. Goodnotes AI Frequently Asked Questions — Goodnotes Support
7. Goodnotes Privacy Frequently-Asked Questions — Goodnotes Support
8. Notability Learn — Ginger Labs Support, June 25, 2026
9. Smart Connections plugin page — Obsidian Community Plugins
10. Using my local LLM with Obsidian — XDA, Feb. 18, 2026
11. Obsidian AI Plugins Setup — Asian Efficiency
12. Why Logseq AI, and how to preserve privacy? — Logseq Forum, May 2023
13. Apple Intelligence Writing Tools Privacy Analysis: What Data Leaves Your Device? — Basil AI, March 6, 2026