The right not to be subjected to AI profiling based on publicly available data
Rapid advances in AI and surveillance are making it trivial and cheap to track and profile individuals based on public and commercially traded data, raising calls for a legal “right not to be profiled.” Commenters question whether such a right is meaningful or enforceable when governments, adtech, and data brokers are deeply invested in profiling, and when data is already widely replicated and hard to delete. Many argue the core problem is profiling and opaque decision systems in general (e.g., credit scoring), not AI specifically, and that meaningful protections would require opt‑in regimes, strict limits on data use, and penalties strong enough to change incentives.
Scope of the problem: AI vs. “just” profiling
- Several argue AI isn’t special: the core harm is profiling itself (by humans, adtech, or data brokers), using both public and private data.
- Others note AI changes things by making surveillance and profiling vastly cheaper and more scalable, turning what used to be rare and labor‑intensive into routine and ubiquitous.
- Some see this as a qualitative shift: “quantity has a quality of its own.”
Surveillance, enforcement, and built‑in inefficiency
- Historically, privacy and “wiggle room” were protected by limits on enforcement capacity; inefficiency functioned as a societal safety valve.
- Automated systems (face recognition, speed cameras, behavioral analytics) threaten to move from ~10% to near‑100% enforcement, effectively making punishments far harsher without changing statutes.
- Multiple comments defend inefficiency as essential to freedom, proportionality, and economic balance.
Rights, regulation, and practicality
- Skeptics see a “right not to be profiled” as unenforceable: once data exists, profiling is technically unstoppable, similar to piracy.
- Others say rights still matter as a legal basis to restrict companies/governments, but enforcement must be against powerful entities, not individuals.
- There is cynicism that the same actors who’d enforce such rights are those most interested in profiling, especially states and large platforms.
- Opt‑out frameworks are criticized as unworkable in complex data/ML pipelines; some argue only strict opt‑in or explicit, narrow allowed-uses can work.
Examples of harmful or dubious profiling
- CRM/AI tools generating personality profiles based on public data are reported as partly accurate but also badly wrong, yet potentially influential for hiring or sales decisions.
- Some suggest such outputs might verge on libel if treated as factual.
- Ad and social media profiles are often wildly inaccurate, highlighting both error and opacity.
- Credit scoring is raised as an existing, opaque profiling system with serious life impact; debate over how “simple” or “nefarious” it is, but broad agreement that lack of transparency and recourse is problematic.
Inevitability vs. mitigation
- Many see ubiquitous AI profiling as inevitable given strong financial and political incentives.
- Proposed “next steps” include: stronger data‑deletion and ownership rights (though their limits are noted), shifting legal liability for holding data, AI literacy, clear labeling of AI‑generated content, and evolving social norms about what is considered acceptable to use or mention.