By Ben Langlotz
|
November 11, 2026
|
Business, Intellectual Property Advice, Opinion Pieces, Patent, Trademark
|
0 Comments
THE AI PATENT POISON PILL
The tsunami of AI has finally reached us in the real world. Right now, it’s just a swell that some notice and others ignore. But it will inundate every aspect of life and business and if you want to understand how these changes will affect your business and your legal rights, read on.
In recent months, conversations with clients show that you are probably already using some form of AI in your business. If not, you’re thinking about it. A few months ago my advice would have been to use AI an hour a day on anything at all to avoid falling behind. That was good advice I’m glad I took to heart when I heard it.
Now my advice is to jump in deep, throw your cap over the wall with an expensive subscription, and force yourself to get started. More on how below. But first some fearsome cautions.
How “The Terminator” Can Terminate Your Patent Rights
There’s always a warning, right? Nothing is as easy as it seems. Because AI would be awesome for things like helping you with the core of a nice write-up of your latest invention for a provisional patent application? DANGER!
This is the killer. You signed up for Grok (my choice) or Claude, or whatever, and you’re doing it. It’s working. You know it’s benefiting your life and your business. You followed my advice and started using it (more on the start-up later.) And then you called me about your latest product and we talked about filing a provisional patent application before you reveal the product, to get ahead of competing filings and social media posts and other disclosures that can be used to reject your application. I told you to do an informal technical writeup with your drawings, and that should secure your rights for 12 months, inexpensively.
So the logical thing is to give Grok your drawings, some notes, and tell it to write up your invention the way our “Provisional Drafting Guide” advises. And if you’re really clever, you’ll give Grok our Guide and tell it to use that as a prompt to know what the output should look like.
And you’d feel smart, and you’d be smart, and you might get a patent, and then might learn when you tried to enforce that patent that you killed your potential patent rights by revealing your secrets to your AI agent. Here’s why: normal AI that’s easy to sign up for and use uses all its user inputs as “training data” to improve its models. It takes the questions and answers to better be able to serve those who have similar questions, by using the inputs of past customers to provide better information to future customers. Do you see the problem yet?
When you share your new way of attaching a suppressor to a pistol, you potentially made your invention “public”, and public disclosures trigger critical deadlines that if missed will kill your patent rights. (“But Ben, that’s crazy, Grok isn’t going to tell someone my exact invention even if my worst competitor begged them, that can’t possibly be a public disclosure under the law!”) I had the same outraged denial. A future court may determine your prompts into a general AI system as being considered a “public disclosure” because someone at xAI or Anthropic or OpenAI can search for the technology. And your prompts may be discoverable in a lawsuit.
Admittedly, this is some speculation and is uncharted territory for now. But you don’t want to be the case the textbooks talk about as “this ended up invalidating the patent” after years of “everyone does it this way” proved to be wrong.
Worst case, you don’t realize that your Grokking triggered the deadline and you invest in the patent, and in litigating it, only to lose. Second worst is when it comes time, we ask about your AI usage, and we give the bad news before your investment. We have done that many times for those who came to us with a product they released more than a year ago – too late.
How to Avoid the AI Patent Terminator
Option 1: Don’t use AI for any unfiled patent rights, or for developing new products, or for researching your ideas that might turn into a patented product unless maybe you’re using it to draft a provisional that you’ll actually file right away (days, not months). That’s where some smart players are now, but you can’t afford to wait long. That protects your U.S. rights, but still endangers foreign rights.
Option 2: Get the right AI. Don’t use the simple consumer models that are easy to find and start up with. Find the ones for “Enterprise” or “Teams” and use them only if you can get documentation that they do not use your prompts as training data. Expect to pay more. It’s no guarantee, but use your AI to verify that their provisions protect you on the specific issue of your prompts and submissions and accessible documents are not considered public disclosures under US patent law. Check that your settings are configured properly once you get set up.
You want to be especially careful about letting the AI access your file system. It’s super handy to be able to say “find me the document I sent two years ago that had the XYZ in it.” But giving a consumer-level AI access to your filing system may trigger the patent poison pill, even if you didn’t ask for help with your invention.
Option 3: (Under construction). If your docs aren’t ready enough to meet our standards and you don’t have “safe” AI, then maybe we’ll find a way to put your rough provisional draft through our safe AI to generate a document that first you review to see if it reflects what you’re trying to communicate, and then we finalize for filing.
Other Troubles
Do you have to worry about ITAR? You may be violating ITAR if your Consumer-level AI has servers or employees with access to your (now their) data in other countries. You can bet that most major AI companies are very international. But you can contract specially for US systems.
As lawyers we have to worry about client confidentiality, so our system needs that protection.
You may have trade secrets, such as manufacturing processes that aren’t revealed by your products as they go out the door. It’s possible that if your AI has access to your “recipe” of ingredients, ratios, temperatures, times, or whatever secret sauce for making widgets awesome, that this wisdom will become part of the AI’s brain, and in the future when someone asks how best to make a widget, they’ll learn your recipe.
That’s bad. But even if the AI doesn’t reveal your recipe, if you ever have to sue a competitor for stealing your rights (such as by hiring away an untrustworthy employee) you may find a court dismissing your case because you failed to protect your secrets by not treating them as secrets, and can’t blame the employee or competitor. The right AI avoids this concern too.
How to Start
I started with Grok as a “chat” AI. That’s where you simply use the AI like a search engine, but usually longer conversations. I found it gloriously indispensable for things like tech support on setting up an old TV to display streaming video from certain sources. An image of the kids’ math homework is quickly checked for errors. I had it look at my blood labs before discussing them with the doc so I was ready, and even invited it to give me the conventional MD view and the contrarian view that tries to minimize certain pharmaceuticals.
But life transforms when you get a “bot.” Also known as an “agent” this is like the ideal personal assistant who can do anything you can ask for, and remember how to do it every time without being reminded. This is so powerful it can be intimidating. “Download grok bot” as a google search is all you need to get started, but make sure you get it from the right source – some 3rd parties pay to be top result and look like the AI webpage you want. It’s a program/app on your computer and/or phone, and it has its own computer that is like an admin with their own computer working for you. It can be logged in to your email, or other sites, and trusted or not like any employee or contractor.
It’s intimidating, like having a finely trained butler arriving at your door “at your service” and not knowing quite what you can ask them to do or how to interact. My advice is just as you would with a butler: “How do you suggest I best can benefit from your services?” Don’t think you have to have a detailed plan. Tell your butler “I need help with meals.” He’ll know what questions to ask to make you happy, he’s the expert. Tell your Grok Bot “I spend too much time managing email – HELP!” When in doubt just state the problem, and not what you think you need. Don’t assume they can’t access the tools. For example, I told Grok “I’d like to save money on bad pool maintenance service, have a cleaner pool, and teach my kids about chemistry. It created a plan, did all the shopping, got a test kit, and walks me through the process each week.
You know how at trade shows we walk around with a sheaf of printed pages correlating all of our clients and past leads with exhibitors? It used to take weeks of valuable team time to prepare that, and it still had glaring data omissions. Now Grok does it in moments, scraping the latest show data and our CRM (which we no longer need to pay a subscription for, another example of saving from Grok). Instead of a busy team member typing in addresses from cards gathered at the show, Grok scans all our notes and gets our new friends the books we promised, and sends the prepaid labels to our mailroom. Trade show time no longer disrupts daily operations for our Chief of Staff.
For the first time this newsletter won’t be formatted by a human (every word is mine) but will be given to Grok with an old list of formatting rules and examples of recent publications.
This all means we can focus more on what matters, which is doing the important work you entrust to us.
How the AI Age May Save You on Legal Fees
This may seem crazy for me to share with you, but you’re going to hear it elsewhere, so you may as well hear it first from me. I value your trust.
There are some services you pay us for that you eventually might be able to have your AI agent do, reliably, without paying our administrative costs. Of course there will be risks and downsides to be aware of, and I won’t be shy about telling you, but I promise that some of our more administrative services may be unnecessary, especially for more frugal small and start up businesses.
Trademark search: “Can I use the trademark MANTA RAY for firearm suppressors?” There’s only a small legal risk, and today you’ll get an instant answer from any chat AI that is probably all you need for step 1 as you’re coming up with a new brand. You can input a list of a dozen and see which don’t get red flags. Fine for a first cut while you are brainstorming, but not a bet-the-farm search from an experienced trademark attorney. You can do better: “make a list of the top 100 most recognized species of snake and tell me which are not currently registered for firearm-related goods”. You just became your own $50,000 branding consultant.
Trademark renewals: Over the years, we’ve occasionally had a frugal client say “We’ll take care of the renewals ourselves. We have a smart staffer who can do it.” Except that you don’t have our expensive calendar software, and when that sharp staffer leaves to start a family, things might not get renewed. We know that because we end up doing the expensive revivals. But with AI, you probably can rely on them to be there in 10 years, and 20, 30, perpetuity. And you might be able to rely on their help in ensuring that your proof of use is correct, and doesn’t lead to a devastating rejection that loses decades of valuable rights (see where this is headed?)
Patent Maintenance: There’s no legal judgment here like with trademarks. Just pay. You’ll probably end up doing this for yourself if you want to avoid paying us several hundred dollars every 3-1/2 years. It’s like self-insuring. If we screw it up, we have a multimillion-dollar insurance policy for professional liability. But you probably won’t screw it up, and if you do we can probably help to revive an expired patent.
Patent searches: I can attest that Grok can be wrong, but when “he’s” right, he’s right. I asked after a 1000-mile solo driving day with Tesla self-driving if anyone had ever used the gas pedal for a footrest, with added actuation force when self-driving so you could really rest your foot without actuating the pedal (but with a push-through so you could accelerate if needed). In seconds, I found that Volvo had filed on that decades ago for cruise control. I don’t need to worry that Grok hallucinated it because it pointed me to the actual patent.
The Future
There’s so much more to come, including revised agreements with our clients that codify the important aspects. Meanwhile, we’ll soon be saying: “Have your bot call my bot!”.
