AI tools are now woven into how we work. They write code, draft copy, design images, and even generate video scripts. But as these tools take on more of the creative load, a hard question emerges. Where does your ownership end and the machine’s authorship begin?
In Episode 62 of Letters of Intent, Pankaj Raval and Sahil Chaudry laid out a practical AI survival guide for founders. They dug into the legal gray areas around intellectual property, plus a serious warning about leaking your own trade secrets. Let’s break down what every business owner needs to know.
The Authorship Dilemma: Who Owns AI Output?
For the first time, humans are not the only ones capable of authorship. AI can generate code, copy, and complete designs in seconds. That creates a genuine legal puzzle.
Here is the core problem, as Sahil framed it. You put your original ideas into an AI platform. That platform was trained on other people’s data. So the output you get back may be based partly on someone else’s content, while your own input may be feeding the system’s future training.
The result is a tangled question of chain of title. Chain of title simply means the documented trail proving you own what you claim to own. With AI in the mix, that trail gets blurry fast.
Consider a real example from the episode. Sahil visited a creators summit where YouTube showed off its Creator Studio tools. These tools help generate concepts, scripts, thumbnails, and titles. Incredibly powerful, yes. But it raises the question immediately: how much of that finished video is truly yours?
The honest answer is that the law is not settled. As Pankaj put it, this is very much the wild west right now. That uncertainty is exactly why founders need to be thoughtful today, before a dispute forces the issue. At Carbon Law Group, we help businesses navigate these questions before they become expensive problems.
Why Chain of Title Matters More Than You Think
You might wonder why this matters if the law is still catching up. The answer is timing. Legal problems with IP almost never show up immediately. They surface later, at the worst possible moment.
Pankaj was blunt about the numbers. Roughly 10 percent of clients ask about AI and ownership up front. The other 90 percent charge full steam ahead. That can work for a while. But the risk sits quietly until something triggers it.
Think about when these issues actually surface. A contractor questions who owns the work. An investor’s due diligence flags a gap. Worst of all, you land in litigation, spending tens or hundreds of thousands of dollars. Suddenly you discover a vulnerability you never knew existed, because you used AI in a way that quietly compromised your IP.
That is the trap. The convenience of AI feels free in the moment. The cost only appears down the road.
So it helps to understand the landscape. There are four main areas of intellectual property: copyrights, trademarks, patents, and trade secrets. Each is complex, with its own rules, and AI touches all of them. This is precisely the kind of foundational review we walk clients through.
Proving Human Input for Copyright Protection
Copyright is where the AI question gets especially tricky. A copyright protects the tangible expression of an idea, not the idea itself. But there is a catch that matters enormously for AI users. The work must be generated by a human.
So what happens when a human prompts an AI and the AI creates the output? Is that so different from using Photoshop? Courts are wrestling with exactly this. As Pankaj noted, a text prompt is probably different from using a pen on a digital tablet. Yet digital tablets already smooth your lines and refine your images. The line is genuinely fuzzy.
Here is the practical takeaway. If your work is primarily AI generated with little human input, you may struggle to copyright it. When you do file, you may need to show the Copyright Office how much human effort went in. Maybe you spent hours refining prompts to reach the final result. That effort could matter in what is shaping up to be a multi-factor analysis by courts and the Copyright Office.
The simplest habit to build is saving your prompt history. Your prompts are evidence of the original ideas you contributed. They help prove the human authorship that copyright law requires. It is a small step now that could protect a valuable asset later.
The Poor Man’s Copyright Strategy for AI Design
Pankaj shared a clever, low-cost tactic for anyone using AI to create designs or images. It starts before you ever open the AI tool.
The idea is simple. Create a crude hand drawing of your concept first. Sketch it out yourself, by hand. Then copyright or officially timestamp that human-made sketch. Because the AI-generated versions become derivative works of your original drawing, you establish ownership over what follows.
How do you timestamp it without filing a formal copyright every time? One classic method is the so-called poor man’s copyright. You mail the sketch to yourself by certified mail. The postal service creates a dated, verified record showing you created the work on that date. If ownership is ever questioned, that timestamp becomes useful evidence.
Filing a formal copyright is always the strongest option. But it is not practical to file for every single image or iteration. That gets expensive and time consuming fast. The hand-sketch approach gives you an affordable middle ground.
Think of it like planting a flag. You mark the human origin of the concept before the AI touches it. That flag anchors your claim to everything the AI helps you build on top of it.
Protecting Trade Secrets From AI Data Leakage
Now for the warning that should make every founder pause. Inputting proprietary company data into a free, public AI chatbot is a serious legal risk.
Here is the danger. Imagine an employee at a biotech company using a public chatbot, and their prompts include the company’s trade secrets. That input could count as public disclosure. And public disclosure can destroy trade secret protection or even kill a patent. One careless prompt can undo years of protected work.
So what should businesses do instead? Pankaj recommends a few concrete safeguards. First, use enterprise-level software, not the free consumer version. Enterprise platforms typically include contractual language protecting your confidentiality, similar to how a business associate agreement works under HIPAA in healthcare.
Second, build a closed system. That means an NDA with the provider and clear internal rules. Redact personal identifying information and client information before it ever reaches the platform. There are even services that layer on top of AI tools to handle this redaction automatically.
Third, opt out of training. Even on free versions of tools like Claude or ChatGPT, you can usually opt out in the settings.
The reality is that we have trusted cloud services like Google Drive and Dropbox for over a decade, and they carry similar risks. The goal is not to avoid AI. It is to limit how much confidential information you expose. We help companies build prompting guidelines and closed systems that keep their secrets safe.
The Litigation Discovery Trap
There is one more risk that catches people off guard. Be very careful about asking AI chatbots sensitive legal questions about your business.
Why? Because whatever you divulge to an AI platform could become subject to discovery in future litigation. Discovery is the process where opposing parties can demand your records and communications. There have already been issues where information a client shared with an AI became discoverable.
Think about what that means. You might type a candid question into a chatbot about a legal problem, describing the situation in detail and maybe admitting a weakness. Later, in a lawsuit, that exchange could surface as evidence. What felt like a private brainstorm becomes a document the other side gets to read.
This does not mean you should avoid AI. These tools are genuinely useful for understanding issues and working faster. The point is awareness. A good rule of thumb is to treat a public chatbot like a public space. Do not say anything you would not want read aloud in a courtroom. For real legal questions about your business, talk to an actual attorney, where confidentiality protections apply.
Building Your AI Strategy the Smart Way
We are still early in the AI era, and the landscape keeps shifting. New tools arrive constantly, including private, on-device models that may soon run entirely on your own computer. The takeaway is not to fear AI or avoid it. These tools are powerful and here to stay. The point is to use them with your eyes open, thinking about the legal implications at every step.
So here is your survival checklist. Save your prompt history to document human authorship. Timestamp your original concepts before feeding them to AI. Use enterprise-level tools with NDAs and redaction for anything confidential. Opt out of training where you can. And never treat a public chatbot as a private or privileged space.
Get these habits right, and you can capture all the upside of AI while protecting the IP and trade secrets that make your business valuable. Get them wrong, and you may not find out until a costly dispute forces the issue.
At Carbon Law Group, we help founders and growing businesses build smart, protective AI strategies. From chain of title and copyright to trade secret protection and enterprise agreements, we make sure your innovation stays yours.
If you want to protect your business as you adopt AI, contact Carbon Law Group today at carbonlg.com to schedule a consultation. Until next time, wishing you the best with your deals and smart risks.