LinkedIn now allows users to opt out of AI training, but the policy excludes direct messages only in Europe
LinkedIn updated its User Agreement and Privacy Policy to explicitly state that it uses publicly posted content to train its generative AI models. The company has not provided the same guarantee as Meta—which publicly committed to excluding private messages from AI training—leaving the status of direct messages ambiguous in LinkedIn’s legal documentation. The opt-out mechanism exists, but the default setting feeds all public user data into AI systems, meaning most users will remain enrolled without taking deliberate action.
How LinkedIn’s AI training policy actually works
LinkedIn’s revised policy permits the platform to use personal data for three stated purposes: improving products and services, developing AI models, and providing personalized services. Users can disable AI training through their account settings, but LinkedIn does not communicate this option prominently during login or account setup. The burden falls entirely on individual users to discover and activate the opt-out.
One significant regional exception already exists: European users’ data is currently excluded from LinkedIn’s AI training due to ongoing regulatory debate in the EU around AI model training permissions. This creates a two-tier system where privacy protections depend on geographic location rather than user choice.
The practical implication is straightforward. If a user posts a professional recommendation, shares industry insights, or publishes a thought leadership article on LinkedIn, that content becomes training data for LinkedIn’s AI products unless they have manually toggled off the AI training setting in their privacy controls.
Meta and X are following similar patterns with regional variations
LinkedIn is not alone in this approach. Meta secured approval to use UK user data for AI training after regulatory review, shifting from an earlier position of excluding UK users. X (formerly Twitter) added an AI training opt-out feature to comply with regional requirements, though like LinkedIn, the default position involves data sharing.
This industry-wide pattern reveals a consistent strategy: platforms obtain permission to use data where legally possible, implement opt-out mechanisms to address regulatory pressure, but maintain opt-in as the default to maximize the volume of training data available. The result is that historical user data—posted before these policies were clarified—has already been ingested into existing AI models, making retroactive opt-out decisions less effective.
The timing problem: opting out applies only to future data
Users who discover LinkedIn’s AI training policy and decide to opt out face a fundamental limitation: the action applies only to data posted after opting out. Any content shared before the user disabled AI training remains in the training dataset. For users with years of LinkedIn activity, this means the majority of their platform history continues to fuel AI model development.
Additionally, aggregation and filtering processes may obscure individual user data within training sets, but the possibility of problematic content generation remains. If an AI model trained partially on LinkedIn data produces output that reproduces biases, misinformation, or other harmful patterns from that training data, the original user who posted the source material bears some indirect responsibility—regardless of whether they later opted out.
This timing gap exists because platforms have financial incentive to delay implementing opt-out features or making them discoverable. The longer users remain unaware of data usage, the larger the training dataset becomes before anyone exercises the opt-out option.
What users should do now
If you use LinkedIn and want to prevent future data sharing with AI training systems, locate the “Data privacy” section within your account settings and toggle off the option for using your information to train generative AI models. This action takes effect immediately but only applies to content posted after the change.
The decision involves a personal calculation about your risk tolerance. Users who post confidential business information, proprietary methodologies, or sensitive client details have stronger reasons to opt out immediately. Users who share only generic professional advice face lower individual risk, though they contribute to broader patterns in AI training data.
Beyond individual settings, consider how your organization handles LinkedIn as a communication channel. If your company policy permits employees to share project details or internal processes on LinkedIn, those practices funnel proprietary information into AI training systems used by competitors and the general public.
If you have expertise in data privacy, AI regulation, or employment law, LinkedIn Daily welcomes practitioner perspectives on this evolving landscape. Visit our write-for-us page to submit a guest post exploring how these policies affect your industry.
The opt-out exists, but it requires active user awareness and deliberate action. Most LinkedIn users will never discover this setting, meaning their data continues supporting AI model development by default.
