Artificial Intelligence (AI) is a powerful tool that can do many good things — like helping doctors in hospitals make personalized health plans. But AI also has a darker side when we use it for surveillance. Systems that use facial recognition and huge databases of personal information can watch people in public spaces without their permission. This can take away privacy and civil liberties, especially if used unfairly. In this article, I will talk about ways people have tried to fix the problem, what didn’t work, what experts worry about most, and which design ideas seem promising.
Some researchers and engineers are already building tools to protect people’s privacy against AI mass surveillance. For example, a system called Fawkes helps individuals “cloak” their photos before they share them online so that facial-recognition software cannot reliably identify them.
Another newer method — Protego — uses a different trick: it transforms a person’s face into a special “pose-invariant” mask, which makes it much harder for face-matching AIs to find that person again. According to its authors, Protego reduces recognition accuracy by a large margin on many AI systems.
On the policy side, many U.S. states (as of 2022) have passed laws regulating how authorities can use facial-recognition technology, trying to prevent unchecked mass surveillance.
Even though tools like Fawkes showed it might be possible to “cloak” faces, they are not perfect. Fawkes and similar methods depend on people proactively altering their photos — which many do not or forget to do. And if someone shares a “clean” (unaltered) photo by mistake, the protection fails.
On the regulation side, rules have often been inconsistent or incomplete. For example, some regions allow certain kinds of AI surveillance under vague terms, which leads to abuse or misuse.
Another failure is that some AI surveillance systems have significant bias. Because many face-recognition AIs were trained mostly on data from light-skinned people, they often misidentify people of color. This not only reduces accuracy — it can lead to unfair targeting or discrimination against certain groups.
Experts warn about several major dangers:
As AI becomes cheaper and easier to use, it could record where people go and what they do — even if they did nothing wrong. That means people could lose their freedom to move or act without being watched. Because many AI surveillance systems perform worse on women or people with darker skin, these groups are more likely to be wrongly flagged. This risks reinforcing unfair treatment and inequality. Often, the people being watched don’t know when or how AI is used against them. There may be no way to check if a system was wrong or misused. If people think they are always being watched, they might be afraid to go to protests, public gatherings, or even normal public spaces. That can hurt democracy and free expression.
From current research and expert suggestions, some of the most effective design features and policies are:
Tools like Protego or Fawkes let people protect their own images — good if enough people use them.
Instead of collecting everything, surveillance systems should only use AI when strictly necessary (for example, investigating a serious crime, not just watching everyone everywhere). Many experts also suggest “data minimization”: keep only what’s needed, delete old data quickly, and prevent wide sharing.
Governments and institutions should make clear policies about when AI surveillance is allowed. Independent audits, public reporting, and the ability for individuals to know/check if they are being watched are important.
To reduce discrimination, AI systems must be trained on diverse data and tested carefully so they don’t systematically misidentify certain groups.
AI has amazing potential to help — for example with personalized healthcare or solving big problems humans struggle with. But when AI is used for surveillance without care, it can harm our privacy, fairness, and freedom. We already have tools and policies that try to protect us, but many of them fall short because of incomplete use, lack of regulation, or technical flaws. Experts say the best path forward is a mix: stronger laws, better oversight, and tools or designs that protect privacy by default. If we build AI carefully and responsibly, we can enjoy its many benefits while still protecting individual rights.
Sources
