A new report reveals that 38 AI companies, data brokers, and other data collectors use manipulative designs to prevent users from opting out of data collection [1].
These findings highlight a systemic effort to bypass privacy preferences. By employing "dark patterns" and fake forms, companies can continue harvesting personal information for AI training and advertising even when users explicitly attempt to stop the process.
The study identified a range of entities engaging in these practices, including defense firms and dating apps [1]. These organizations utilize deceptive user interfaces to make the opt-out process either ineffective or intentionally difficult to navigate [1], [3].
Some companies have gone beyond confusing layouts to implement entirely fake forms [2]. These pages mimic legitimate privacy request portals but do not actually trigger the removal of user data from the company's systems [2]. This tactic ensures that the flow of data remains uninterrupted while providing a veneer of compliance with privacy expectations.
The report suggests these tactics are used to maintain the volume of data available for commercial purposes [1], [2]. Because AI models require massive datasets to function, the incentive to keep collecting information often outweighs the commitment to user privacy controls.
These deceptive practices are occurring across various websites and applications where opt-out forms are presented to the public [1], [3]. The use of such patterns suggests a calculated approach to user retention and data acquisition that prioritizes corporate growth over individual digital autonomy.
“38 AI companies, data brokers, and other data collectors use manipulative designs to prevent users from opting out”
The discovery of fake opt-out forms indicates a shift from passive 'dark patterns'—such as confusing menus—to active deception. This suggests that current self-regulatory privacy frameworks are insufficient, as companies may prioritize the data requirements of AI training over legal or ethical obligations to user privacy.
