LinkedIn is training AI on your data by default

LinkedIn just pulled a classic Big Tech move: quietly feeding your professional history and posts into its AI models before even updating its user agreement.

First spotted by 404 Media, users found a new setting in their account privacy tabs that was toggled on by default, granting the platform permission to train generative AI on their personal data. TechCrunch and Ars Technica noted that LinkedIn updated its privacy policy only after the new setting went live, with general counsel Blake Lawit confirming formal contract updates taking effect on November 20.

Here is the real catch: opting out only protects your future data. Whatever LinkedIn already ingested into its AI models is staying there.

A double opt-out and a data warning

According to LinkedIn's updated FAQ, the training pipeline swallows posts, articles, usage frequency, language settings, and interactions with AI features. Those datasets are used to train in-house models for writing assistance and post recommendations, but may also be shared with corporate parent Microsoft and its Azure OpenAI service.

LinkedIn claims it uses privacy-enhancing techniques to strip identifiable personal info, but its own documentation admits a glaring risk: if you paste personal details into an AI prompt on the site, that data could theoretically show up as generated output for someone else down the road.

Stopping this requires jumping through two distinct hoops. First, you must navigate to Settings, head to Data Privacy, click on Data for Generative AI Improvement, and turn the toggle off. However, that only applies to generative models. If you want to block your data from being used in non-generative machine learning tools like feed recommendation or content moderation, you have to find and fill out a completely separate Data Processing Objection Form.

Regulators step in

Unsurprisingly, LinkedIn never enabled this automatic training in the European Union, the European Economic Area, or Switzerland. The threat of massive GDPR penalties was enough to keep European profiles off the menu from day one.

Users in the UK were not so lucky initially. However, after the UK Information Commissioner's Office (ICO) raised concerns, LinkedIn confirmed on September 20 that it paused AI training on UK user data.

The scramble mirrors recent backpedaling by Meta and X. Meta paused EU data ingestion after complaints from privacy group Noyb, and X agreed to permanently stop scraping EU user data for its Grok model following action from the Irish Data Protection Commission.

Why it matters

We think this episode lays bare how desperate Big Tech has become for clean, professional training data. As public web scrapes run dry, LinkedIn's repository of career history, industry commentary, and resumes is pure gold for foundational model builders like Microsoft.

That said, LinkedIn handled this poorly. Sneaking a default opt-in toggle into user dashboards before changing the terms of service looks like a deliberate end-run around informed consent. Splitting the opt-out into a toggle and a separate web form feels uncomfortably close to a dark pattern, and declaring that past training cannot be undone presents users with an unfair fait accompli.

If you use LinkedIn outside the EU or UK, you should go disable the toggle and submit the objection form immediately. More broadly, the contrast between European protections and the rest of the world makes it crystal clear: voluntary self-regulation is dead, and only hard regulatory teeth will stop tech giants from taking whatever data they can reach.