The Valdosta cyclist incident, where a rider was hit by a car, has the legal world buzzing again about how old laws handle new technology. The hot topics are AI work product and legal privilege, because let’s face it, AI tools are everywhere in our practices now. And frankly, a ton of bad advice is floating around about what’s actually protected work product when an AI is involved.
Key Takeaways
- Work product protection for AI-generated material is a non-starter unless a lawyer specifically directed the AI’s work as part of a concrete litigation strategy.
- Just using an AI doesn’t make its output privileged. The lawyer’s own judgment and strategic thinking are what earns the protection, not the tool itself.
- Judges are starting to look closely at filings made with AI help, so you have to disclose its use if it affects the facts or arguments you’re presenting.
- Using AI means you’re still on the hook for your ethical duties of competence, confidentiality, and supervision. Mess that up, and you can kiss work product and attorney-client privilege goodbye.
Myth 1: Any Document Touched by AI is Automatically Protected by Work Product
This is just flat-out wrong, and it’s a dangerous belief. The work product doctrine from Hickman v. Taylor is about protecting materials prepared for litigation. The whole point is to stop opposing counsel from getting a free ride on your hard work. The key part is “prepared… in anticipation of litigation.” If you use an AI to generate a document or a research memo without a lawyer giving it specific marching orders for a particular case strategy, it’s almost certainly not getting work product protection. Imagine your firm uses a tool like DISCO AI to scan years of client emails for a general compliance check, long before a lawsuit is even on the horizon. The AI is doing complex work, sure, but its output is just an administrative review. It’s not tied to an imminent or ongoing legal dispute, which is what the doctrine requires. An AI’s involvement doesn’t magically turn a routine business analysis into a protected legal strategy. For the documents to even have a shot at protection, a lawyer has to be in the driver’s seat, setting the AI’s parameters and using its output to build a specific litigation plan. That hands-on, strategic direction is everything.
Myth 2: AI-Generated Content is Always Attorney Work Product if a Lawyer Presses “Generate”
A lot of lawyers seem to think that if they’re the one typing the query into the AI, the output is automatically their work product. That’s a ridiculous oversimplification of the doctrine. Remember, work product doctrine distinguishes “ordinary” work product (discoverable if the other side shows a big need) from “opinion” work product, which gets almost total protection because it contains a lawyer’s thoughts, theories, and mental impressions. An AI can’t have ‘mental impressions’, it’s a machine. When a lawyer uses a tool like Thomson Reuters AI to summarize case law, the initial output is just a fancy organization of public information. Unless that lawyer carefully crafted the prompts, fed the AI specific facts to analyze, and then heavily rewrote the output to inject their own unique legal arguments, that draft isn’t opinion work product. It might be ordinary work product, at best. And even that protection is flimsy if the other side can argue they can’t get the info any other way. What matters is the human lawyer’s intellectual contribution. If the AI is just pulling data or spitting out boilerplate, it’s a real stretch to claim it contains a lawyer’s strategic thoughts on the Valdosta cyclist case. Without that human element, it’s just high-tech data processing.
Hit while cycling?
Most cyclists accept the first offer, which is typically 50–70% less than what they actually deserve.
Myth 3: AI Use Doesn’t Need to Be Disclosed in Legal Filings
Candor toward the tribunal is one of the most basic rules of our profession, and that rule applies to how we prepare our documents. The idea that you can use AI behind the scenes and not tell anyone is dying fast, with courts and bar associations killing it. We’re already seeing judges in jurisdictions like the Northern District of Georgia issue standing orders that require you to disclose when you’ve used AI to do your legal research or draft your pleadings. The issue is accountability and the basic integrity of the court system, not just whether the AI’s output is accurate (though that’s a huge worry too). When an AI “hallucinates” a fake case which has happened more than once, the attorney who filed that document is the one on the hook. Georgia Rule of Professional Conduct 3.3 is clear: you are responsible for the accuracy of what you file. Hiding your AI use when it formed a material part of your filing’s factual or legal basis is a great way to get sanctioned and torpedo your own credibility and your client’s case. Being transparent about AI assistance for research and drafting is quickly becoming non-negotiable.
Myth 4: Client Confidentiality is Automatically Maintained When Using AI Tools
This is another spot where a simple misunderstanding can get you into serious ethical trouble. Our duty to protect client confidences is fundamental, spelled out in O.C.G.A. Section 15-19-4 and Georgia Rules of Professional Conduct Rule 1.6. Just dumping client information into a third-party AI without reading the fine print on its terms of service is asking for a bar complaint. Many public-facing AI tools, especially the free or general-purpose ones, might be logging your inputs to train their models, which could be a catastrophic breach of confidentiality. Think about the lawyer for the Valdosta cyclist. If they plug sensitive medical details or settlement numbers into a public chatbot to help draft a demand letter, they could be handing that data right over to the AI company. It’s easy to see how that data could be used to train future AIs or, even worse, get exposed in a data breach. You have to do your homework on any AI vendor. Read their data privacy policies, check their encryption, and see if they claim any rights to the data you put in. Using a tool without knowing those answers is a clear violation of your duty of confidentiality. Firms need to be using enterprise-grade AI with solid data protection contracts or, even better, running private, on-premise AI models so client data never leaves the building. Keeping client secrets safe requires this level of vigilance.
Myth 5: AI Tools Are a Substitute for Attorney Supervision and Professional Judgment
Some people talk about AI like it’s a magic wand that can handle complex legal work on its own. This is a dangerous fantasy that completely ignores the actual job of a lawyer. AI tools are powerful assistants, but they can’t exercise judgment, they don’t get nuance, and they certainly can’t build an empathetic relationship with a client. The Georgia Rules of Professional Conduct, Rule 5.1 (on supervising partners) and Rule 5.3 (on nonlawyer assistants), apply just as much to your AI tool as they do to your paralegal. You are responsible for every piece of work that goes out under your name, whether it was drafted by a first-year associate or an algorithm. If an AI-generated brief has a fatal flaw, misreads a statute, or misses a key argument for the Valdosta cyclist’s claim, that’s on you, the supervising attorney. The AI is just a tool, like Westlaw or Word. It needs constant human oversight, critical review, and the application of your professional judgment. Relying on an AI’s raw output without checking its work and refining its content is professional negligence and it cuts against the very heart of what we do. Lawyers have to review, edit, and validate everything an AI produces to make sure it fits the case strategy and meets our ethical duties. AI is becoming part of the job, but using it properly means knowing its limits and the ethical traps that come with it. Staying on top of this stuff requires constant attention and a solid grip on our core professional principles.
So what’s the key for AI content to get work product protection?
It all comes down to whether a lawyer directed the AI’s work for a specific litigation purpose. The output has to reflect the lawyer’s own strategic thinking, not just be a data dump the AI created on its own.
Can using a public AI tool for research blow up client confidentiality?
Absolutely. If you feed confidential client info into a public AI and its terms of service let it keep or use that data, you’ve likely violated your ethical duties under Georgia Rules of Professional Conduct Rule 1.6.
Do I have to tell the court I used AI in a filing in Georgia?
There’s no statewide rule yet, but it’s coming. Federal courts like the Northern District of Georgia already have standing orders demanding it. Given your duty of candor, disclosing AI use is the only safe and ethical choice.
Can an AI’s output ever be “opinion work product”?
Almost certainly not, at least not the raw output. An AI doesn’t have mental impressions or legal theories. The only way it gets close is if a lawyer so heavily edits and infuses the AI’s text with their own strategy that it becomes their own thought product.
What’s the main ethical duty when I’m using AI tools?
Your duties of competence (Rule 1.1) and supervision (Rules 5.1 and 5.3) are front and center. You have to understand the tool’s limits and you must review and take responsibility for everything it produces, just like you would for a human assistant.
