A pattern has settled across the software industry. A platform launches or expands an AI feature. A policy update appears in a changelog or settings page. The new default is that your data contributes to AI training — and opting out requires finding the right menu and turning something off.
This is not a conspiracy. It reflects the economics of building AI products: proprietary data from real users is valuable, and making contribution the default with opt-out available for those who look for it is an effective acquisition strategy.
Here’s where the toggle is on the platforms most likely to hold your data — and what it actually controls.
Atlassian (Jira, Confluence, Jira Service Management)
What’s happening: Starting August 17, Atlassian will use content from Jira tickets and Confluence pages to train its AI products — Rovo and Rovo Dev — by default.
What’s included in training:
- Jira issue titles, descriptions, and comments
- Confluence page bodies
- A separate metadata category (aggregated usage signals) that non-Enterprise customers cannot opt out of
How to opt out (all tiers, for in-app content):
- Go to admin.atlassian.com
- Navigate to Security → Data contribution
- Turn off in-app data collection before August 17
Who can act: Organization administrators only. Individual users cannot independently opt their own content out.
What isn’t covered: The metadata category — de-identified aggregated usage signals — remains on for Free, Standard, and Premium tiers. Only Enterprise-tier customers can disable metadata collection.
Sub-processor note: Atlassian’s AI pipeline routes through sub-processors including OpenAI. Atlassian’s data residency offering covers data at rest but does not cover the AI training pipeline — data stored in EU-based residency may still be processed in the United States for AI training.
What’s happening: LinkedIn trains its generative AI features using posts, comments, articles, and other user-generated content by default. The setting has been live for some time, but many users haven’t found it.
How to opt out:
- Open LinkedIn Settings & Privacy
- Go to Data Privacy
- Find “Data for generative AI improvement”
- Toggle it off
Scope: LinkedIn states the opt-out applies to future content and removes your data from training pipelines. Content that has already been used to train deployed models cannot be retroactively removed — the opt-out is forward-looking.
What remains: Opting out of AI training doesn’t affect LinkedIn’s other data collection for ad targeting, analytics, and its standard service operations.
Dropbox
What’s happened: In early 2024, Dropbox updated its AI feature terms in ways that indicated user content could be processed by third-party AI providers. After significant public backlash, Dropbox clarified that AI features were opt-in and that content was not used to train AI models without user action.
Current status: AI features in Dropbox — including Dash (AI search) and AI-generated summaries — are described as opt-in. When users enable these features, their files are processed by AI systems. The concern is less about training and more about which files get processed for feature delivery.
How to check your settings:
- Go to dropbox.com → Account Settings → Security
- Review which AI features are enabled in your account
- Check Connected apps for any third-party AI integrations you may have authorized
If you enabled any Dropbox AI feature without fully reading the terms, verify what data processing those features involve.
Canva
What’s happening: Canva updated its terms to allow uploaded content to be used to improve AI features. The primary concern is images and designs uploaded by users potentially informing Canva’s AI generation models.
How to opt out:
- Go to canva.com → Account settings → Privacy & data
- Look for controls labeled “AI training” or “Data usage for AI”
- Toggle off any AI training settings present
Why it matters for businesses: Canva is widely used for client presentations, marketing materials, and brand assets. Company-confidential designs may fall within the scope of AI training if the organization’s Canva accounts haven’t been audited for this setting.
Meta (Facebook and Instagram)
What’s happening: Meta uses publicly visible content on Facebook and Instagram — posts, photos, captions, comments — to train its AI systems. It has faced regulatory pressure in the EU, where it paused this practice under scrutiny from data protection authorities.
How to opt out on Facebook:
- Settings → Your Facebook information
- Privacy rights for people in your region → More options
- Find AI training controls and submit the opt-out form
How to opt out on Instagram:
- Settings → Account → Privacy
- Navigate to AI training or data use settings
EU status: Meta paused AI training on EU user data in 2024 following regulator objections. Whether and under what conditions this resumes is subject to ongoing proceedings with EU data protection authorities.
What isn’t covered: Private messages and non-public content are not used for AI training under Meta’s disclosed policies. The opt-out applies to public content.
What’s happening: Google’s privacy policy states it may use publicly available content to improve AI products. For content stored in Google services (Drive, Photos, Gmail), the specifics vary by product and account type.
For Google Photos: Google uses image data to improve AI features including search, face recognition, and content suggestions. Within Google Workspace, admin-level controls govern how data is used for AI feature improvement.
How to check:
- Go to myaccount.google.com → Data & privacy
- Review “Web & App Activity” settings — this governs whether your activity personalizes Google AI features
- For Workspace accounts: Admin console → Account settings → Workspace data use
Limitation: Google’s controls are distributed across multiple settings menus. Even with careful configuration, the full scope of what Google uses for AI improvement across its product portfolio is difficult to audit comprehensively. The clearest protections available to individuals are within Google Workspace enterprise settings, not the consumer account controls.
Microsoft (OneDrive, Microsoft 365, Teams, Copilot)
What’s happening: Microsoft has stated it does not train its foundational language models on consumer data from Microsoft 365 for commercial accounts when appropriate enterprise settings are in place. However, when Microsoft Copilot features are enabled, they process your file content to generate responses.
How to check:
- For enterprise accounts: Work with your Microsoft 365 admin to review Copilot data settings and the settings under Admin center → Settings → Org settings → Copilot
- For consumer accounts: account.microsoft.com → Privacy → Privacy dashboard
- Determine which Copilot features are active and what their specific data handling terms say
Encryption note: Microsoft holds the encryption keys for content stored in OneDrive and Microsoft 365 consumer and standard commercial plans. When Copilot or AI features process your files, that processing occurs within Microsoft’s infrastructure on content it can read. Enterprise customers with specific compliance configurations (Customer Key, Microsoft Purview) have additional controls.
Notion
What’s happening: Notion AI processes the content of your pages to generate summaries, answer questions, and assist with writing. Notion’s terms allow it to use interactions with Notion AI to improve AI features.
How to check:
- Go to Settings → AI settings in your Notion workspace
- Review which AI features are enabled and whether content is sent to third-party AI providers
- Review Notion’s current data processing terms for the specific scope of AI training
Third-party note: Notion uses external AI providers including OpenAI to power its AI features. When Notion AI processes your pages, that content is sent to those providers under data processing agreements. The agreements constrain secondary use, but the data leaves Notion’s own infrastructure.
The Pattern Worth Recognizing
Looking across these platforms, several things are consistent.
Defaults favor the provider. With rare exceptions, the default is that your data contributes to AI improvement unless you act. The opt-out exists, but you have to know to look for it.
Opt-out settings are fragmented. There is no standard location, no universal preference signal, and no central dashboard. You need to audit each platform independently on its own settings.
Controls are imprecise. Even after opting out, platforms typically retain the right to use content that was processed before the opt-out date. The exact scope of what a given opt-out covers is not always clearly disclosed and often requires reading the specific policy language.
Enterprise gets more control. Across Atlassian, Microsoft, Google, and others, Enterprise-tier customers consistently receive more granular opt-out controls than individuals or small business customers. Privacy protection, in these systems, scales with what you pay.
What You Can’t Audit Your Way Out Of
Even a careful opt-out audit leaves some exposure. The controls above apply to disclosed training use — they don’t address the AI processing that happens as part of actively using features (summarization, search, content suggestions), which involves your data being processed even when that processing is not labeled “training.”
More fundamentally, the opt-out approach places ongoing maintenance on users. When a platform updates its terms — which happens regularly — new data categories or use cases may require finding and activating a new opt-out. Settings you configured last year may not cover categories introduced this year.
For files where you genuinely need confidence that content won’t become training material — personal documents, private records, sensitive photos, health information — the reliable answer is not a setting found in a menu. It is a platform that is not in the AI training business to begin with.
daftei never trains third-party AI on your files, and there is no opt-out setting to find because there is no training to opt out of. Files are encrypted with AES-256 at rest and TLS 1.3 in transit. The platform is GDPR and CCPA compliant, does not sell data, and does not run ads. Storage starts at 5 GB free, with unlimited storage on the Pro plan at $5.99 per month. What you store there stays there.