Unchecked AI rollout threatens privacy rights

AI portends a vast shakeup in the way even the most mundane business processes are conducted.

Nothing is more humiliating than being scrutinized in a store based on suspicion of shoplifting.

Unfortunately, this humiliating scenario unfolded in recent years at Rite Aid stores nationwide. A facial recognition system, powered by an artificial intelligence algorithm exhibiting inherent bias against minorities, compelled retail clerks to confront individuals flagged as shoplifting suspects.

The practice continued for nearly eight years in what Rite Aid said was an experimental program deployed in a “limited number of stores.” Nevertheless, the impacts of the practice were wide-ranging, subjecting an untold number of innocent shoppers to suffer embarrassment, often in front of friends and relatives. Many had to defend themselves against unfounded accusations.

“Rite Aid’s reckless use of facial surveillance systems left its customers facing humiliation and other harms,” said Samuel Levine, the director of the Federal Trade Commission’s Bureau of Consumer Protection, when announcing a 54-page complaint against the pharmacy chain.

Rite Aid stands as a noteworthy example illustrating the repercussions of deploying artificial intelligence algorithms without rigorous testing and vetting to prevent unintended consequences. 

Many other enterprises will experience similar issues if they adopt AI without meticulous reviews.

The allure of the technology may distract them from recognizing potential pitfalls. But it should go without saying: Do not allow your organization to fall into the same trap.

My experience within the industry gives me a sense of foreboding about its ability to self-police. Instances of failure, exemplified by the Rite Aid case, will become more common and prompt regulatory authorities to take stringent measures in the near-term as AI technology continues to encroach on fundamental privacy rights. As data collection grows exponentially through various means, including the Internet of Things, the threats only loom larger.

We have at least come a long way from the wild-west scenario that reigned little more than five years ago, when widespread data breaches and a general lack of accountability was rampant. Individual privacy and security was a mere afterthought as emerging technologies were embraced and introduced to the public with little regard to basic protections. One need only refer back to the troubling Equifax breach that exposed credit assessments of what was conservatively estimated to be 150 million people.

With those episodes as a backdrop, a movement toward implementing stringent privacy safeguards is well underway. The FTC, which previously took action against Rite Aid, is now initiating measures to bolster privacy safeguards for children. Reports indicate that these efforts involve limitations on tracking by various services, including social media apps, video game platforms, toy retailers, and digital advertising networks. This marks only the beginning of these initiatives.

Now, as AI takes hold and portends a vast shakeup in the way even the most mundane business processes are conducted, it is time to redouble data protection efforts. There needs to be a sense of urgency on a national data privacy initiative. 

We see some baby steps. One year after the launch of ChatGPT, President Biden issued an executive order aimed at ensuring that the widespread introduction of AI is "safe, secure, and trustworthy." The order is designed to strengthen government oversight of technology to prevent adverse impacts, including the reporting of “red team” tests that detect hidden flaws in the technology.

As the deployment of technology accelerates, growing concerns about privacy have become evident. Recent measures by the European Union — including the European Data Protection Board's restriction on Meta processing personal data for behavioral advertising — alongside efforts by state legislatures to protect personal privacy, underscore a rising apprehension surrounding the rapid advancement of AI.

Much of that anxiety comes from opacity by major AI players about the training data used to inform their respective models. Indiana University researchers have already detected flaws in current systems, workarounds that bypass what developers say are safeguards that prevent the access of private information.

The revelations by researchers should send out ear piercing alarms that personal information is at high risk, and, left unchecked, could be the largest threat to personal privacy ever witnessed.

Data protection compliance is a legal requirement. But more importantly, it is an ethical imperative. There is one sure fire way to doom the reputation of your company — to be fast and loose with personal data.

Just as the U.S. Constitution lays out certain fundamental civil rights and liberties, so too should a new digital bill of rights be codified to protect online activities and personal data.

The European Union’s General Data Protection Regulation and California Consumer Privacy Act are both a good start. Rather than create 50 sets of state privacy rights regulations, the U.S. requires a national standard.

Assuring data privacy should not be a cat-and mouse game of developers trying to stay one step ahead of regulators. In essence, people have the right to know what data is being collected, where it is stored and who has access to it. And this should not be in some multi-page document couched in legalese, but easily digested, in the same form and method consumers can currently request their own credit reports.

Stricter data protection rules could disrupt revenue streams at some social media concerns that count the sale of personal information into their business models. Some services now offered for free may have to initiate modest fees in the name of data protection. That may have to be the ultimate trade-off for the public.

Just as developers must have a data-privacy consciousness, so too should users. A degree of responsibility falls on the public to monitor what data to share. That means employing virtual private networks, encrypted messaging apps and demanding a level of transparency from those who collect data. 

Ultimately, the method to protect personal data is through a partnership based on trust between developer and user, with regulatory authorities behind the scenes setting the standards.

Scott Allendevaux is senior practice lead at Allendevaux & Company, a data protection agency.