I’m 27. I look both ways when I cross the road. I don’t have hundreds of dollars spare to spend on professional legal assistance. I don’t have a Will.
I do, however, draft Wills for people all the time. As a lawyer, it’s part of my job.
And with the hype surrounding the capabilities of ChatGPT and AI, it’s a job I fear I may not have forever.
So, I decided to put a theory to the test.
Could I use OpenAI’s ChatGPT to create a legally binding Will for myself? How easy would it be to replace me? Could I become the first human with an AI-generated Will?
First, I set some ground rules. Obviously, given my qualifications, I have a bit of knowledge most people won’t have, so I needed to exclude that.
I’m a lawyer in New Zealand, not the USA. I don’t know US law. So, the rules were:
- No Googling.
- No attempts to educate the AI.
- I needed to actually create and physically execute the Will.
- The Will needed to be valid in both the USA and New Zealand.
Here’s how it went:
Want to talk to me about this article? Email me: [email protected]
In This Article
Getting Started
I log in to ChatGPT and open with the only question I realistically can: How can I create a will?
The AI provides me with a helpful answer broken down into steps.
The last two steps relate to signing the Will and storing it. The software can’t do those for me (yet).
A bonus step it tells me is that I ought to seek the advice of a lawyer. Ha. I’m using you to avoid doing just that!
Step 1: Assets and Beneficiaries
I decide to follow the steps ChatGPT suggests in the order suggested, starting with my assets.
Unfortunately, I can’t think of any major assets I own. Of the assets I do have, I think I want to leave most, if not all, of them to my partner, Emily*.
I ask the AI for guidance:
At this point, I’m already a little spooked.
So far, the output from ChatGPT is almost verbatim what I tell clients regularly. Right down to the breakdown of types of assets to consider (although I note it hasn’t considered the topic of debts).
Moving forward, I settle on giving Emily everything except for my gaming PC. My PC will go to my brother (he will love it).
I let the AI know my wishes. I ask him (her?) to remember this information for later, hoping that it might make the “Okay Siri, now write my will” part of this process a bit easier.
I also make sure to say please. It feels wrong to do otherwise.
The AI kindly suggests, again, that I consult with a lawyer.
I am starting to get concerned that I will reach a point where it tells me it can’t help me anymore.
I’m aware OpenAI is working on adding new “safety mitigations” with each update, and I’m unsure that it will keep wanting to help me with something so critical (so critical that I had to go to University for five years to do it).
Regardless, I continue onwards and decide I would like to give something to a charity I have an affinity for. I ask ChatGPT to jot that down as well.
Step 2: Executor
With my assets and beneficiaries settled, I move on to the AI’s next proposed step:
Choosing an executor, who will apparently “be responsible for carrying out the instructions” in my Will.
Unsure who to pick, I ask the software for guidance:
So, I need to pick someone trustworthy to manage my assets and fulfill my wishes.
I’m trying to save money here, so a professional executor is out of the question.
I’m tempted to ask the AI if it will be the executor for me, but I know I’ll be disappointed by the answer.
So naturally, I’m going with Emily. She meets all the criteria.
The AI also suggests I choose a backup executor. In case the double-decker bus crashes into us at the same time, I suppose.
My brother makes sense for this backup role, and I feed it into the machine.
Apparently, I’m supposed to let Emily and my brother know I’ve made this decision. I guess I’ll do that by sending them a copy of the article once it’s published.
I ask ChatGPT for any final tips, and it prompts me on my funeral arrangements. That’s a good point and something I hadn’t considered.
Upon reflection, I want to be cremated. It would be kind of cool to have my ashes put in fireworks (I think I recall that being a thing?).
But honestly, I haven’t given this much thought, and I’d rather defer to Emily, as she’s the one who will ultimately be sorting it all out once I’m gone. I let the chatbot know.
Step 3: Writing the Will
With all that out of the way, I’ve covered everything ChatGPT asked of me, and now it’s time for the moment of truth: I ask the AI to write my Will and wait patiently for its response.
Damn.
I was worried about this. The AI doesn’t want to play ball.
In some ways, this is a good thing; my job is safe for another day.
But I’ve come so far, and I don’t want to give up here.
I plan to move our conversation over to the OpenAI playground, which I understand is less strict than ChatGPT. I ask ChatGPT for a summary of my data which it happily provides in a bullet-pointed list, and I’m about to switch over to OpenAI when I have a cheeky idea.
The AI refuses to draft my “real” Will.
But will it draft a hypothetical example Will for me, assuming I assure the AI it’s totally hypothetical and totally just an example?
Success.
That’s just the first half, but it gives me full Will from start to finish using the information I had provided. I copy it into a Word document, fill in names and addresses, and send it to the printer! Now it’s time to get this thing signed.
Step 4: Signing
I am now wholeheartedly determined to be the first person in the world with a legally valid AI-generated Will. I put ChatGPT’s clear indication that I’m working in a soon-to-be-dead profession out of my mind.
With my document printed, I ask my coworkers to be my witnesses. Then I pause again. Can coworkers be my witnesses under US law?
There is only one place to check:
Perfect. I explain to my coworkers what I’m doing, and I eventually find two of them happy to participate.
They witness me signing, and the deed is done. Signed and executed.
The Finished Product
Is My ChatGPT Will Legally Valid?
In New Zealand, undoubtedly, this document is my last Will and testament, and it will give effect to my wishes as I’ve expressed them.
Under Section 14 of the Wills Act 2007, a Will is valid as long as it’s in writing, signed, and witnessed by two non-beneficiaries.
I can say with authority that if I die tomorrow, Emily will have to apply for probate to distribute my estate, and my AI-Generated Will will get a big red stamp of approval from the New Zealand High Court.
In America? I’m not an expert. But according to Berk Law Group, in Arizona, at least, yes. The requirements are broadly the same as in New Zealand.
Takeaways
The accuracy of the AI is mindblowing to me. It got 99% of what it said to me during this process entirely on point. It gave almost identical advice to that which I have given clients before.
However, it’s still relatively shallow. My wishes are not complex. Online Will generators with templated forms have been around for a while, and my Will is probably something one of those could have achieved.
Furthermore, the Will isn’t perfect. Yes, it’s valid, and yes, it will give effect to my wishes, but there are a few minor ambiguities that I would edit if I didn’t think putting my life in the hands of a computer for an online article was more important.
I’ve made a few jokes about ChatGPT taking my job. And honestly, if my job was writing basic wills for people with simple wishes – yes, I would probably be out of that job soon. However, thankfully, most lawyers, including myself, do many other things, including things that AI is far from being able to replicate (yet).
Conclusion
I pasted the above in ChatGPT and asked it to write me a conclusion. Here it is:
In conclusion, I have attempted to use OpenAI’s ChatGPT to create a legally binding Will.
The AI provided me with a helpful step by step guide, breaking down the process of creating a Will into assets and beneficiaries, choosing an executor and writing the Will.
Despite the AI’s repeated suggestions to consult a lawyer, I was able to determine my assets, beneficiaries and executor using the AI’s guidance.
However, the AI was not able to physically execute the Will or store it, which are critical steps in the process.
This experiment highlights the limitations of AI in the legal field, as well as the importance of seeking professional legal assistance when creating a Will.
I don’t think we’re quite there yet – James, a human.
2 Responses
That is fascinating! I’m an attorney in NY (USA) and I have been playing around with that type of scenario too. And yes, it’s an easy trick to go into hypothetical mode with ChatGTP or I’ve had luck asking it to write the script of a “play” and gotten the results I wanted.
I’ve drafted estate planning documents for my clients for over 20 years. And I work with a company that helps attorneys draft documents for clients’ estate plans. I’ve read about Harvey AI being used for estate planning. Do you know of any other AI systems for estate planning?
Hi Joe. Thank you for your comment.
I don’t know any AI systems for estate planning that are “quite there” yet, in being something I’d want to use in day-to-day practise, but I am very excited to see how this space develops over the coming years.