Atlanta AI Surveillance: Privacy Rights in 2026

Evelyn, a dental hygienist with over 20 years on the job, never thought her career would get derailed like this. It was early 2025. She was helping a patient at a busy Buckhead dental office when a faulty chair just gave way. She fell hard, sustaining a nasty back injury. The initial diagnosis came back as a herniated disc, setting her on a path of extensive physical therapy and the possibility of surgery. Her employer’s workers’ comp carrier, a big national company known for playing hardball, started its investigation right away. Evelyn had no idea how much they’d lean on AI-powered surveillance, which started to blur the lines of her privacy and raised serious questions about just how far this tech can go in Atlanta WC claims. Mr. Thompson, the insurance adjuster, seemed to know an unnerving amount about Evelyn’s day-to-day life, things she thought were totally private. He mentioned her shopping trips to the Kroger on Piedmont Road, seeing her at her son’s baseball games over at Chastain Park, and even brought up her gardening, a hobby she was doing mostly for therapeutic reasons. A creeping discomfort set in. “How do they know all this?” she asked her attorney, Sarah Miller, a partner at a well-known downtown Atlanta firm that specializes in workers’ comp. Sarah wasn’t surprised. She’d been seeing more and more of these kinds of situations over the last year as insurance companies rushed to adopt advanced AI tools. Mr. Thompson’s vague explanation was that the AI just “analyzed publicly available data” and flagged “anomalies.” Sarah knew surveillance itself was nothing new in workers’ comp, but AI was adding a whole new layer of complexity. Old-school surveillance meant hiring a PI to physically watch a claimant. AI, on the other hand, could scrape and combine huge amounts of data from social media, public records, commercial databases, and even camera feeds accessible to the public. This change gave insurers new tools but created serious problems for claimants’ privacy rights. Luckily, Georgia law isn’t silent on this. O.C.G.A. Section 34-9-17, for example, lays out an employee’s rights and responsibilities, which indirectly sets the guardrails for what an investigation can and can’t do. The State Board of Workers’ Compensation (SBWC) has its own rules that get more specific about admitting evidence, including surveillance. “The key thing,” Sarah told Evelyn, “is whether their surveillance, AI or not, actually proves you aren’t hurt or are faking it. It’s not about whether they can *find* you on camera. It’s whether that footage is relevant, accurate, and they got it legally.” She explained that while Evelyn’s gardening might look bad at a glance, like if the AI flagged her lifting a heavy bag of potting soil, it could be used to argue her back injury wasn’t as bad as she claimed. The real issue is when the AI starts making assumptions based on bad or incomplete data. Maybe Evelyn was gardening, but she was using a special long-handled tool to avoid lifting, a nuance an algorithm would completely miss. AI surveillance works by using algorithms to spot patterns in activity data, looking for inconsistencies between what a claimant says their limits are and what they’re seen doing. So if Evelyn said she couldn’t lift more than five pounds, but the AI tagged a video of her carrying a twenty-pound grocery bag, that’s a red flag for the insurer. But what if the AI didn’t see her husband carry the heavy bags? Or that she used a cart? Or that she was in agony for two days afterward? This is where the whole concept of AI investigation starts to fall apart. AI has no human nuance. It can’t grasp context, understand pain, or tell the difference between doing something once in a pinch and being able to do it all day long. And you don’t even know where they’re getting this stuff. Insurers often use third-party data brokers that scrape information from all kinds of public and semi-public sources. We’re talking social media profiles, public records, and even location data you can buy commercially. While a lot of that data is technically “public,” having an AI scoop it all up and analyze it puts the idea of a reasonable expectation of privacy to the test, even when you’re out in public. The Georgia Court of Appeals has been pretty consistent that people have an expectation of privacy even in public, especially for personal activities not easily seen by just anyone. During a pre-hearing conference with the SBWC, Sarah put Mr. Thompson on the spot about his AI surveillance. “What algorithms are you running? What data are you feeding them? How are you correcting for bias in the data or the AI’s misreading of human activity?” she asked. Mr. Thompson got defensive, hiding behind “proprietary software” and “trade secrets.” Sarah had heard it before. It’s a common tactic. Insurance companies treat their AI methods like they’re top secret, which makes it almost impossible for an injured worker to see the evidence being used against them. That lack of transparency is, in her professional opinion, a massive flaw in how AI is being used in workers’ comp right now. Sarah’s argument was that the AI’s findings were thin and circumstantial, lacking the kind of direct proof required under Georgia law. She got ready for a hearing, building her case around O.C.G.A. Section 34-9-102, which covers hearing procedures. Her plan was to have Evelyn’s treating doctor, Dr. Chen from Emory Orthopaedics & Spine Center, testify about the real-world specifics of her injury and her actual medical limitations. Dr. Chen’s testimony would supply the medical context the AI, by its very design, was incapable of understanding. A critical piece of Sarah’s strategy was hammering the distinction between public and private. Yes, Evelyn was in public at Kroger and Chastain Park. But her home and her private doctor’s appointments were off-limits. If it came out that the AI surveillance had crossed that line by accessing private messages or snooping on her property, it would be a clear violation of her rights, and a violation of the Georgia Code of Professional Conduct for any attorney trying to use it.

At the hearing, Sarah went to work dismantling the carrier’s AI “evidence.” She brought in an expert to testify about how limited current AI is when it comes to accurately judging human physical ability and pain. She showed how the AI actually flagged Evelyn going to a physical therapy appointment, an activity that proved she was trying to get better, not faking an injury. The algorithm, with its simple binary logic, just saw “activity” and couldn’t figure out the purpose. Sarah then presented evidence of Evelyn following her doctor’s orders to the letter, like using assistive tools for chores and carefully pacing herself. The AI’s supposed “anomalies” were just Evelyn trying to live a normal life within her very real limitations. The administrative law judge (ALJ) on the case, Judge Reynolds, listened to all of it. Having presided over hundreds of workers’ compensation cases, he understood the realities of physical injury and recovery way better than any algorithm ever could. In the end, Judge Reynolds sided with Evelyn. He found that the insurance carrier had a right to conduct surveillance, but the AI-powered evidence they presented was not convincing. It didn’t prove she was faking or exaggerating her injury. He pointed out that the AI’s conclusions completely missed the context needed to overcome the solid medical evidence from Dr. Chen and Evelyn’s own consistent testimony. The judge made it clear that the burden of proof is on the party claiming fraud, and that flimsy inferences from an AI, without solid human verification, just don’t cut it. The decision was a huge relief for Evelyn. She could finally get back to focusing on her recovery without constantly looking over her shoulder for a robot spy. For Sarah, it was a big win, not just for her client but for every injured worker who might face the same fight. The case showed that even with fancy AI, the core principles of Georgia’s workers’ comp law and a person’s right to privacy still hold. Insurers have to play by the established rules of evidence, and claimants can and should challenge the accuracy of whatever an AI spits out. We’re going to see more AI in the legal world for sure, but it will always be measured against basic human rights and hard-won legal precedent.