By Tyler Balding:
Employment law has long served as a proving ground for how legal ideas harden into doctrine. Because it sits at the intersection of economics, institutional power, and social policy, employment law often absorbs contested assumptions early and then carries them forward long after their origins have faded from view.
The most familiar example is at-will employment. Today, it is treated as a background principle so fundamental that it is rarely interrogated. Yet legal historians widely agree that at-will employment rests on a historical misstatement that hardened into doctrine through repetition, convenience, and institutional inertia.
That history matters now more than ever, because it closely parallels a growing risk associated with artificial intelligence: hallucinated assertions, once relied upon and repeated, can quietly become law.
The danger is not that AI gets things wrong; the danger is what happens when no one checks. This is especially true in a field that already knows how easily unverified assumptions can become binding rules.
The At-Will Doctrine: How an Employment-Law Error Became Orthodoxy
The modern at-will doctrine can be traced with unusual precision to a single source: Horace Gay Wood, a New York lawyer who, in 1877, published a treatise titled Master and Servant. Writing with confidence rather than caution, Wood asserted that the common law had long recognized employment relationships as terminable at the will of either party unless a fixed term was expressly stated. He presented the proposition not as a contested view, but as settled law.
Courts proved receptive. Trial courts cited Wood. Appellate courts cited those trial courts. Soon, courts were no longer citing Wood at all, they were citing one another. Within a remarkably short period, Wood’s formulation stopped looking like an argument and began to look like a description of reality. By the early twentieth century, the at-will rule no longer required explanation. It was simply “the law.”
The difficulty, as later historians and scholars painstakingly demonstrated, was that Wood’s account did not accurately reflect the common law he claimed to summarize. Many of the English and American cases he cited did not stand for at-will termination at all. Earlier courts often presumed year-to-year employment or required cause for dismissal, particularly in skilled trades and long-term service relationships. The historical record was uneven, contextual, and fact-dependent. Wood’s rule was not.
But by the time those inconsistencies were exposed, it was too late. Employment law, perhaps more than any other field, rewards rules that are easy to administer. Once courts began relying on the at-will doctrine to resolve cases efficiently, they had little incentive to reopen its foundations. Each citation became another layer of insulation. The doctrine’s authority no longer depended on whether it was historically correct, but on the simple fact that it was already being treated as correct. That is the critical mechanism by which legal fiction becomes legal reality, not through conspiracy or bad faith, but through repetition, convenience, and institutional trust.
AI Hallucinations Operate the Same Way
Large language models are not malicious. They do not “lie.” They predict text based on patterns. When source material is ambiguous, incomplete, or conflicting, AI fills the gaps in a way that sounds authoritative.
The risk emerges when those outputs are relied upon without verification, repeated by others, embedded in briefs, policies, pleadings, or articles, and later cited as if they reflect settled fact.
At that point, the hallucination no longer looks like an error. It looks like consensus. This is precisely the same dynamic that allowed at-will employment to take hold. Confidence substituted for accuracy. Repetition substituted for proof.
When an AI Hallucination Actually Entered a Court Order
Until recently, concerns about AI hallucinations in law were largely hypothetical—warnings about what might happen if fabricated authority slipped through the cracks. That line has now been crossed.
In Shahid v. Esaam, 918 S.E.2d 198 (Ga. Ct. App. June 30, 2025), the Georgia Court of Appeals vacated a trial-court order after concluding that the order relied on nonexistent case law that had been cited by counsel. The appellate court expressly noted that the trial court’s written order had incorporated bogus authorities, cases that did not exist in any reporter or database, and imposed sanctions in connection with their use. The error was ultimately corrected. The order was vacated. The fictitious cases did not become binding precedent, but would have with on more simple oversight.
But the significance lies in what happened before correction: hallucinated case law crossed the institutional threshold and entered an operative judicial order. That is the precise point at which legal fiction stops being theoretical. Employment law teaches us that not every error is caught immediately and that some survive long enough to shape doctrine before anyone realizes what happened.
Institutional Incentives Make the Problem Worse
Employment law developed in response to industrial efficiency, labor mobility, and economic pressure. Courts favored simple, repeatable rules. At-will employment fit that need. Modern institutions face similar pressures with AI: speed over verification, cost savings over primary research, plausibility over provenance.
AI text is especially dangerous because it does not announce its uncertainty. A hallucinated case citation can look indistinguishable from a real one. A fabricated historical “fact” can sound identical to a well-sourced one. And once that material circulates, particularly in employment policies, handbooks, or litigation templates, it acquires legitimacy simply by being written down. It isn’t at it’s core a technology problem; but rather a process problem.
Why Employment Lawyers Should Be Especially Concerned
Employment law is uniquely vulnerable to this phenomenon. It is precedent-driven, policy-sensitive, and often shaped by broad generalizations rather than narrow holdings. Doctrinal shortcuts, like at-will employment, tend to persist precisely because they are useful.
If AI-generated errors enter employment-law practice through briefs, internal guidance, training materials, or template policies, they may not be challenged immediately. They may simply be repeated. Over time, they can reshape how rules are understood, even if no single case openly endorses them. That is how employment law has always evolved, for better and for worse.
The lesson of at-will employment is not that courts are careless or that technology is dangerous. It is that systems reward convenience, and once a convenient rule is accepted, it rarely gets revisited. At-will employment survived not because it was correct, but because it was useful. AI hallucinations will survive for the same reason, unless institutions impose discipline.
Conclusion: Employment Law as a Warning and a Test Case
AI can be an extraordinary tool for employment lawyers: drafting, research, and pattern recognition. But it must be treated the way lawyers treat junior associates, helpful, fast, and never authoritative without review.
Primary-source verification is not optional in this new paradigm. Institutional norms must treat AI output as a starting point, not an endpoint, especially in a field where assumptions can harden into doctrine for generations.
Employment law shows us how easily legal fictions can become foundational rules. It also shows us how difficult those rules are to unwind once they take hold. The Georgia case demonstrates that AI hallucinations are no longer hypothetical. They have already entered court orders, if only briefly. History suggests that the next error may last longer. Further, a comprehensive survey has never been done to see where such errors may have already worked their way into orders unnoticed.
AI does not introduce a new risk; it accelerates an old one. The question is not whether AI will hallucinate… it will. The question is whether lawyers, who already live with the consequences of historical misstatements, will stop, verify, and ask whether the thing everyone is repeating is actually true before it becomes the next at-will doctrine.
If you would like more information about this article or speak to one of our employment attorneys, please contact Carey & Associates, P.C. at info@capclaw.com or call 203-255-4150.
