One of the coolest portions of this, was having it suggested to put the contract through ChatGPT. I (immediately) scoffed at the idea, that it would get it wrong. To prove me wrong, said person put it through their ChatGPT account and an analysis came out.
This one: https://chatgpt.com/s/t_694201aae96c8191b11272903a3bd2b7
Mea culpa. It was a LOT better than I expected. It caught the major, major red flags that JR and I caught. For a first look, just to see if you should keep talking? Not bad.
So. Not bad. But why don’t I think you should use it and/or rely on it? I’ll explain below.
Let me be clear, btw – I’m not going to talk about ethics or morality or the legality of anything done. Furthermore, I should note that I’m a Luddite mostly (in both the historical sense and the modern one); about this tech. I don’t know it, I am just using the output here as it stands.
So… Moving on. We’re skipping the portion where they covered the red flags, those were all correct and highlighted the stuff JR and I both caught (and missed some others, but again, not the focus of the discussion).
Mind you, this AI chat is based off the prompt, ‘should you sign it’, so it’s a very b&w decision. Maybe other prompts might have other data (probably, but I’ll get to that).
This is the section that summarises stuff, starting with ‘what’s good’:
Having read through and analysed the rest of this, things like ‘no upfront fee charge’ statement when taken in context of saying it’s good is disingenuous at best.
Royalty percentages can be stated but terrible (see this contract and SOOOO many others). Also, note that there aren’t statements at all about what a good royalty rate should be (or any discussion of that at all).
Publisher indemnifies author – this I don’t actually think is true that other contracts don’t do that. In fact, I see this quite often.
Semiannual accounting – is for the AUTHOR’s Future Earnings side. Their royalty reports are quarterly. So hallucination / wrong logic.
Acknowledges AI – and says they’re only going to follow the law. That’s it. It’s a non-clause.
The final line – ‘good intentions’? This paints a picture that whoever wrote this MEANT to be good. Such language gives the publisher a pass, and is just wrong.
Moving on to ‘What’s Missing’
Alright, so here’s a document I put together a while ago. It’s my own cheat sheet for when I look at a contract, to see what might be missing: Contract Review Guidelines
Let’s go over this.
- audit rights? Great!
- reversion triggers for non-performance. Agreed. Maybe you could prompt to see what these might be… maybe not.
- publication timeline – yes! Otherwise, they can technically hold a book for 2-4 years, not publish it and you are stuck. This sometimes happens in trad pub who delay releases due to constricted publication schedules, etc.
- marketing commitments – this is not standard. Never has been, really. If they are there, they’re super vague.
- caps on recoupable costs – this should not even be asked for. Recoupable costs should be redlined out at all. Why are you suggesting that this should be included at all?
- Sublicensing approval rights – this is suggested for certain rights in the model contract by the author’s guild. I’ve seen and not seen these rights, but it certainly is something I’d negotiate for. Decent catch.
- Author approval on adaptations – I can’t say since the contracts I’ve seen either don’t include the rights, have them set-up under sublicensing approvals (or no approvals) or just not included. Might be an overreach ask.
- Estate protections – what does this even mean? I need more data before I comment.
- Force majeure protections – for the author? Or the publisher? Because most force majeure stuff is on the publisher, though on the author side might not be bad. Not all contracts have them, but it’s not a bad addition.
- Material breach – yes. We discussed this. Lines up with reversion triggers.
- Explicit out-of-print clause – what kind? Examples here would be great. I’d be interested to see what this looks like.
Next comes a section that explicitly says, don’t sign. Good!
Alright, this is what they want to fix if you wanted to go through this. I’m NOT going to go through this section point-by-point, we’ve discussed many of these things (or will). I will say, with a contract this big, limiting your changes to 10 points seems silly.
What ChatGPT missed or didn’t go into detail on
- Reversion rights, often called ‘out of print’ rights – this is a tricky one. Old contracts had ‘out of print’ defined as books not being available for sale. With digital works, that doesn’t work. These days, the best out of print clauses use a combination of units sold and revenue generated in an annual term. Units by themselves aren’t enough because of free promos, but revenue generated can work by itself.
- Sublicensing rights – (Missed by ChatGPT) there should be some discussion of ‘third party’ rights, and defining these 3rd parties such that the rights cannot just be resold to another company owned by the publisher for $0. Also, if they do want to use it for a subsidiary, setting forth negotiation protocols.
- Royalties – we had a long discussion about royalties and royalty rates. The AI doesn’t highlight how long audiobook royalty rates are, and certainly isn’t bringing any nuance to the discussion. Also, if you’re negotiating, you might want to start considering how many tiers or thresholds you want, if you’re using unit sold or $’s made and at what levels. All things I don’t think the AI would be helpful with, since everyone’s case is different. What I can get from a publisher will be different from a debut author.
- Advances – AI never once asks for an advance. This is, as mentioned in the previous blog post, a good way of making publishers work, especially if you are acquiring a large advance. A $1k advance won’t make a difference to big publishers. A $10k advance starts looking bad. A $100k means they will work for it, if they can.
- Publishing timelines – are mentioned, more explicit ways of reverting rights if rights aren’t used would be great.
- Author copies – not mentioned. Getting some would be good.
- Survivability / Bankruptcy / Sale of Company – again, not mentioned. Needs to be in there, so they can’t declare bankruptcy and then resell all assets (your IP / contract) to another company. Mind you, there’s actually discussion that some of these terms might not matter, but better to have them.
My Problem With this AI Output
Set aside the random hallucinations, especially when it messes up the Future Income side with the royalty payments of the main contract. Or adding terms that aren’t explicitly standard, because it thinks it should be. Or the lack of nuance in some of the suggestions (and thus negotiation that might occur, though again – don’t negotiate a contract like this. Walk away).
Through this, I’ve been picking at what they’re suggesting, what should be done or not done, what can be asked or not. What’s a hallucination.
I’m able to do that because I’ve spent a bunch of time reading model contracts, other contracts and my own. I have a little bit of contract law knowledge, and ask questions / go to seminars / workshops on this when I can (at Author conferences). I have an in-built set of knowledge.
To ask the questions I have, I needed to do the hard work.
Someone who is coming to use ChatGPT who hasn’t done the hard work is not able to ask the questions or find the problems that I did. They would then try to negotiate a contract using the suggestions, but would then miss (quite literally) all the things above that I pointed out isn’t there.
Some of which are quite important.
Maybe you could find it, explicitly asking more questions from the AI. You could go through multiple rounds, over and over again, trying to figure out how to reword things or what might be missing.
However, if you’re then using the AI to ‘learn’, how much of what you’re learning is wrong? Remember, most of this was right – but a small portion was hallucinated or wrong. How do you tell if it’s hallucinated without doing the real work of finding sources outside and reading all that?
If you’ve gone that far, why involve the AI at all?
So, is the AI doing a good job at telling you to run from this contract? Yes.
Would I trust it to help me negotiate or pick out all the other issues? No.
Last Point (and why there’ll be one more blog post after this)
If the AIs were the ONLY way for someone to get advice, without doing the hard work or paying a lawyer their $500-1000 fee, I might hold my nose and say, go ahead.
However, there are numerous free resources on model contracts out there. There are numerous individuals (tap into your author community, ask around!) who will look at and help people on contracts. There are full on groups and associations who will do a full review for free, or free with membership.
My next post will be a list of resources. But, yeah… use them. Learn it yourself. Hire an Entertainment & IP Lawyer just before you sign, if necessary.
Heck, if you have an agent – this is what you pay them for. (Again, know it yourself is best too!).
Recent Comments