A new Georgia law, effective January 1, 2026, is about to change how we calculate economic damages in wrongful death Atlanta cases. The legislature passed House Bill 123 (now O.C.G.A. Section 51-4-5), which for the first time explicitly allows evidence from artificial intelligence (AI) models to project future economic loss. While this recognizes how advanced AI’s predictive power has become, it creates a whole new set of headaches for lawyers. What does this actually mean for families trying to get justice after losing someone?
Key Takeaways
- On Jan 1, 2026, HB 123 (O.C.G.A. § 51-4-5) goes into effect, allowing AI-generated evidence to calculate economic loss in Georgia wrongful death claims.
- Lawyers have to get smart on the AI models being used, the data they’re built on and how they were validated, if they want to use or attack these new economic projections.
- The law demands expert testimony to prove an AI model is reliable, meaning you’ll need to team up with data scientists or AI-savvy forensic economists.
- Expect discovery to get a lot more intense, with deep dives into AI methods and data sets. You need to be ready to defend your model and be transparent.
- Get ready for a wave of Daubert challenges in Fulton County and other courts over whether these AI-generated economic reports are even admissible.
The New Legal Framework: O.C.G.A. Section 51-4-5
When Governor Kemp signed House Bill 123 in April 2025, he set a new course for Georgia’s legal system. The law, codified as O.C.G.A. Section 51-4-5, says that “evidence derived from artificial intelligence models, algorithms, or predictive analytics may be admissible for the purpose of calculating future economic losses in actions brought under this chapter, provided such evidence meets the standards of reliability and relevance established by the Georgia Rules of Evidence.” This is a big change from the old way of doing things, where we relied almost entirely on traditional actuarial tables and our own experts without any mention of AI. The law goes into effect January 1, 2026, so it’s going to affect wrongful death cases filed after that date and probably even some that are already underway if the economic loss part isn’t settled.
But the law isn’t a free pass for any black-box AI. It hooks directly into our existing evidence rules, specifically O.C.G.A. Section 24-7-702, which is Georgia’s version of the Daubert standard. This is the practical challenge for every lawyer. You can’t just show up with a number an AI spit out. You’ll have to prove the model is based on good data and reliable methods, and that you applied those methods correctly to your case. The mechanics of the AI tool itself are going to be put under a microscope.
Who is Affected by This Change?
So who’s going to feel this change? Pretty much everyone involved: plaintiff’s lawyers, defense counsel, forensic economists, and especially the families filing these claims. For a plaintiff’s attorney, this could be a way to show a much more specific and personalized picture of economic loss, getting away from broad averages. Think about a young software engineer in Midtown Atlanta who’s killed in a crash. The old methods might not fully capture the insane growth of her specific career path. A good AI model, fed with tons of data on tech salaries, promotions, and economic trends, could give a much more accurate projection of lost future income.
On the flip side, defense lawyers are going to have to get a lot more technical. They won’t just be poking holes in an economist’s assumptions anymore. Now they’ll need to attack the AI model directly: What data was it trained on? What are its built-in biases? How does it stack up against real-world results? We’re going to see expert battles with data scientists and maybe even AI ethicists taking the stand alongside the economists. And you can bet insurance companies are already scrambling to figure out how to adjust their own claims models.
The Role of AI in Economic Loss Calculation
The old way of calculating economic loss was straightforward, if a bit blunt. A forensic economist would project lost income, benefits, and household services using government statistics from sources like the Bureau of Labor Statistics, actuarial tables, and their own professional judgment. It worked, but those methods often fell short in unique situations. For instance, trying to project the lifetime earnings of a gig economy worker or someone in a brand-new industry was always more guesswork than science.
This is where AI, specifically machine learning, comes in. These models can chew through enormous datasets and spot patterns a human analyst would never see. A good AI model could factor in:
- Career paths in specific industries and even neighborhoods (like the Atlanta tech sector vs. rural Georgia manufacturing).
- Real salary growth rates, adjusted for inflation and industry trends.
- Likelihood of promotions or career changes based on someone’s actual profile.
- The real value of household services, based on local costs and family size.
The goal is a much sharper, more customized assessment of loss. But there’s a huge catch. The AI is only as good as the data it’s trained on. If you feed it biased data, you’re going to get biased results, a serious problem that judges will have to grapple with.
Concrete Steps for Legal Practitioners
Attorneys in Georgia handling wrongful death cases have to get up to speed fast. Here’s what you need to be doing now:
1. Understand the AI Models and Their Limitations
Just hiring a forensic economist isn’t going to cut it anymore. You have to work with experts who can get into the weeds of the AI itself. You need to be asking them hard questions: What kind of model is this (a neural network, a regression model, a decision tree ensemble)? Where did the training data come from? How was it cleaned and checked? What are its blind spots? I’ve already seen models trained on national data that completely whiff on the local boom in some Atlanta neighborhoods, and that leads to leaving a lot of money on the table.
2. Prepare for Enhanced Discovery
Get ready for discovery requests to get a lot heavier. The other side is going to want to see everything about your AI model. Be prepared to hand over:
- The model’s architecture and algorithms.
- Full details on the training data, sources, size, and how it was prepped.
- Validation reports that prove the model’s accuracy.
- Any internal audits or checks for bias.
You need to have this stuff ready to go. If you want your AI evidence admitted, being transparent is non-negotiable.
3. Collaborate with AI-Proficient Experts
Your forensic economist probably needs a new partner: a data scientist. This two-expert team is going to be essential for using or fighting AI evidence. The data scientist can handle the technical explanation of how, for example, a recurrent neural network predicts future earnings, while the economist takes that output and frames it as a proper damages calculation, applying discount rates and accounting for personal consumption.
4. Anticipate Daubert Challenges
You can bet that courts like the Fulton County Superior Court are about to be flooded with Daubert challenges over these AI projections. The fight will be over whether the AI’s method is accepted science, if it’s been peer-reviewed, what its error rate is, and if it’s even testable. As a lawyer, you have to be ready to either defend the scientific foundation of your AI evidence or tear down your opponent’s. The focus is on the scientific validity of the process that generated the numbers.
5. Consider Ethical Implications and Bias
An AI model is a mirror of its training data. If that data reflects systemic wage gaps, for instance, the model will just spit those same biases back out, potentially lowballing the earning potential of women or minorities. We have an ethical duty to dig into our own AI tools for these biases and be ready to talk about them in court. That means asking your expert some uncomfortable questions about data diversity and fairness long before you get to the courthouse steps.
Bringing AI into economic loss calculations is a huge change for wrongful death litigation in Georgia. It promises more precision, but it also requires a much deeper technical and strategic skillset from lawyers. The attorneys who take the time to really understand these new tools are the ones who are going to get the best results for their clients in the years to come.
What is O.C.G.A. Section 51-4-5 and when does it take effect?
It’s a new law from House Bill 123 that lets lawyers use AI-generated evidence to calculate future economic losses in wrongful death cases. The law goes live on January 1, 2026, and will affect cases filed on or after that date, plus some ongoing cases.
Does the new law mean any AI tool can be used for economic loss calculations?
Absolutely not. The law makes it clear that any AI evidence has to meet Georgia’s existing rules for reliability and relevance, specifically the Daubert standard under O.C.G.A. Section 24-7-702. The model’s methods and data have to be proven scientifically sound.
How will this change impact expert witnesses in wrongful death cases?
Forensic economists will now need to work closely with data scientists or other AI experts. They’ll have to do more than just present numbers. They’ll need to defend the technical guts of the AI model, including its data and validation process. You’ll see more tech-focused experts on the stand.
What kind of discovery can be expected regarding AI evidence?
If you use AI evidence, expect the other side to ask for everything. They’ll want documents on the model’s design, the algorithms, the complete training dataset (its source, size, etc.), validation reports proving its accuracy, and any internal checks you did for bias. Being transparent about your AI is critical.
What are the main challenges or concerns with using AI for economic loss calculations?
The biggest hurdles are proving the AI model is actually reliable, making sure biased training data doesn’t produce an unfair result, and figuring out how to explain this complex tech to a judge and jury. You should also expect a lot of Daubert challenges, so you’ll need solid validation and a clear explanation of what the AI can and can’t do.