Mistral Patent for “Code implemented tool calls”
patentsgazette.uspto.gov
patentsgazette.uspto.gov
Software patent are a scourge of the software industry. Patents are designed to protect costly research; simply having an idea is not costly (but it makes in medical research for example). All that software patents do is creating a minefield that hinders competition.
For software Copyright is a far better instrument. Let the one with best implementation win... That's where the cost is: Implementing, testing, shipping, maintaining. Protect that.
Sorry for the rant.
Edit: Spelling
But this is minimally protectable, and you can copy the idea just not the exact implementation. Copyright does not have a 'doctrine of equivalents' equivalent. It also supports independent creation. Copyright also does not cover functional elements. You cannot copyright an idea, only the expression. So maybe you can elaborate more on why your point makes sense?
If that idea had cost (say) a billion dollars (like in some medical research or physical research), there absolutely is value in granting a temporary monopoly in order to encourage such research in the future.
I don't think a "pay-to-play" system works. There's already trade secrets that allow corporations to invest lots of money in protecting things that aren't even inventions. You just seem like you're trying to figure which patents are good or bad. But if a patent is widely infringed, by the same kind of logic, it must have tremendous value as an invention.
Beyond that, it's a trivial batching pattern. Any remotely competent systems engineer developing an agent would immediately reach for it, it's the most basic tool in the toolbox to deal with latency. If Mistral "invested a lot of money in discovering" this, man I'm in the wrong line of work.
1. Researcher makes something
2. Employer patents it using researcher's name
3. Researcher gets a new job
4. Researcher makes something at new job vaguely similar to #1
5. Old employer sues for patent violation
By attaching their name to the patent, it is easy to prove they were aware of it.
The boogieman that software patent advocates raise is that IP theft and free riding will destroy America if we don't have patents. But literally nobody is creating software to get a patent on it, and software patents don't stop anyone from replicating software.
Instead software patents are just a way for mediocre lawyers to live parasitically off the tech industry.
Imagine a world where we were so worried about bullying that we created a department to monitor all speech on the Internet, and you had to run your speech past lawyers for approval before saying anything. And you could be sued for saying anything by anyone if you didn't get lawyer approval. That would be a world with dramatically less bullying than the Internet today. AND a much worse world with much less communication. AND in that alternate world, the lawyers would argue that there would be more bullying if we put that ridiculous system to an end. And they would be right that there would be some more bullying, AND it would also be right that that degree of bullying would be worth it to have free speech.
Similarly with patents. Yeah, on the margins there might be more IP theft. SO WHAT. This system right now sucks.
America needs to learn how to evolve our system of industrial production. Our society of lawyers and paperpushers is going to get annihilated in the next ten years at this rate.
I do not like it one bit, but that does seem to be state of the software world right now.
Actually novel implementation is protected, paid the author's bills, and was dedicated to the public domain recently - no massive corp involved.
Parents point seems to be about that if this author of Slug wouldn't have patent it in the first place, then in general that work would have had an larger impact in the software industry, compared to this situation where the author did patent it.
Case(s) in point: all the comments on HN who independently came up with the same algorithm but couldn't publish it as FOSS because they found a patent covering it.
Us programmers are essentially arranging blocks. And if you arrange them in a special way you can get your stylus to follow them in amazing ways.
I don’t see why that’s very different than arranging atoms.
At the lowest level, software is always represented physically in the medium. It’s concrete.
I'd argue that how they should have been, but the reality is quite the opposite. The patent framework is all about novel ideas rather than costly research, even though ideas aren't any valuable anymore - the easy stuff that is cheap to reproduce once the idea is known is already discovered.
Multimedia codecs and similar stuff are pretty patent-worthy (costly research) but their value is an exponential function of their adoption, and hence the unencumbered one always wins. Same with basically everything else.
That's where the cost <was>
I think with LLM, implementation winning is moot now.
Interesting physical real world problems are NP hard. That's why we have patents. Software? You can work backwards from the requirements to multiple solutions almost always. Which is why patents on software are stupid.
Copyrights are originally to protect authors because an author might create a few dozen works over a lifetime. Of which a few mich actually be worth anything. Copyrights were to protect his income and allow him to leave an inheritance behind.
Copyrights for software is stupid because most of it is mechanically generated and it goes old like last weeks fish. Quote: The source code for Windows 3.1 who cares.
However this completely overlooks the fact that the true value of software, and theoretically what we should protect, is not the code, but the functionality it provides. Copyright explicitly does not cover functional elements -- that is the realm of patents -- and really only covers code to the extent that it has "expressive elements." Technical creativity (novel algorithms, data structures, architecture etc.) is expressly excluded.
One non-obvious and absurd implication of this is that any code that can be shown to be purely dictated by technical constraints is not protected -- https://en.wikipedia.org/wiki/Abstraction%E2%80%93filtration... via https://en.wikipedia.org/wiki/Computer_Associates_Internatio.... -- yet a simple rewrite using equivalent code constructs, like using a while-loop instead of a for-loop, can be enough to keep a snippet of code from infringing the original. (I'm oversimplifying of course, but principally this is how the law is.)
If you look back at significant copyright-related cases for software now (e.g. CA vs Altai above or Oracle vs Google), you'll realize they have twisted themselves into knots trying to apply laws intended for expressive creativity to issues that were essentially about technical creativity.
So copyright covers pretty much the literal code, and the compiled software as a "derivative work", and nothing more. Things only happened to work so far because reverse engineering and re-implementation was expensive.
But as comments in this thread are pointing it out, AI is making it very stark that the code itself matters very little. We're just now realizing how ineffective copyright has always been for code.
I have no hopes that we will figure out an appropriate IP framework for software, so I expect people will move towards other things like patents, trade secrets and trademarks. Which have their own problems, but at least they already exist and are more suitable than copyright, especially in the age of AI.
Clearly this is an attempt to prevent similar patents from being weaponised against them in the US.
No one cares about such stuff in Europe.
Thunderbirds are Go!
I'm glad my servers are operating in a deep undersea cave.
https://en.wikipedia.org/wiki/MP3#Licensing,_ownership,_and_...
Fraunhofer is a research institution. They didn't really do licencing, that's why they never went after anyone.
Then Thomson joined the venture (wouldn't you know it, French!) and the fun with litigation got started
Not trying to protect Frauenhofer there at all, if they didn't smell opportunity they wouldn't have teamed up with Thomson after all... But I think the trend is very clear. Germans are generally a lot worse at the litigation game then the french. Historically that makes sense, but it's definitely surprising that it still holds true in modern times
Fraunhofer didn't do licencing before joining up with Thomson. That's why - back in the 90s - there were no litigation anywhere even though everyone ignored their patents.
Then Frauenhofer joined up with Thomson and that's where your understanding joins up with reality: now the Frauenhofer does licencing and sues anyone that's not paying them.
My whole comment was a technical tidbit - maybe meaningless from today's perspective... But the previous view interpreted a forward thinking strategy with the monetization of mp3 that just wasn't there
This is pretty common when a company exclusively licenses a patent: they demand that the patent holder sue anyone using the tech who isn't them, so that they actually benefit in the market from the exclusivity. Without that enforcement, the money they were paying for exclusivity would be for nothing.
Thomson just made this kind of arrangement in the context of a non-exclusive license — getting Fraunhofer to start legally enforcing their patent rights at all, so that Thomson and whoever else licenses Fraunhofer's patents could benefit in the market from having licensed the patents, by ensuring that non-licensees aren't getting to use the patented tech for free.
In the US, such an arrangement is unnecessary (to the point that it might be causing a miscommunication in this conversation!), since patent holders tend to legally protect their patents all on their own, suing non-licensees for use of the patented tech even when there are no licensees demanding to "get the benefit they paid for." In Europe, patents aren't sued over by default, with enforcement instead only beginning when the patent is licensed and the licensee demands enforcement so that they can benefit.
They don't do this sort of thing... oh wait, they do.
But we are still wrong to judge their character by their own actions, because... evil French people made them do it?
Why do we assume that Mistral is a "good guy" protecting themselves? I'm not sure they necessarily are. Could just be trying to grab any ground they can.
I mean, maybe they are the "good guy", but I don't think we know either way.
The submitter needs to be identified, so get a friend/representative to submit if you want to maintain anonymity.
Or an Inter Partes Review!
There are “workflows” that are very similar. Ie I can invoke a Python function that doesn’t exist and the exception gets handed off to an LLM to make it work by making that function exist.
This is functionally similar from what I can gather. Invoke a tool that doesn’t exist, we’ll make it exist for you.
It’ll probably get a patent issued, though, the patent office seems asleep at the wheel in the tech space at least.
Whether a patent is valid is tested in court.
[1] https://learn.microsoft.com/en-us/agent-framework/agents/cod...
Tool calls were immediately obvious, and everyone and their dog was doing their own hacky version of it (today we call it "harness engineering") for a good year before OpenAI wrapped it up in an API.
It’s one of those kinds of patents i assume I’m violating every 100 lines of code I write.
Patent was applied for 04.03.2026. langgraph-codeact _shipped_ this capability in February.
[0]: https://platform.claude.com/docs/en/agents-and-tools/tool-us...
[1]: https://developers.openai.com/api/docs/guides/tools-programm...
I only have a rudimentary understanding of how neural networks work, but I wonder if rather than “understanding” what “novel” really means to humans, an LLM would be most likely to agree that something was novel based on having seen that specifically referred to as novel in its training data.
So that if you give it an example of something that already exists, but which was very recently invented at the point in time when the LLM was trained, and you ask “is this a novel idea?” that because it had several sources in its training data describing that idea as novel, it would say “yes that’s a novel idea”. Whereas what we really meant was to ask it if someone else had already thought of this thing prior to us right now in this later moment.
And then on the other hand, even if something was “novel” at the point in time when the LLM was trained, perhaps we would fare better to ask it “has anyone thought of this?” rather than asking if the idea is “novel”? And that even though it considers the idea novel in a way it would also be able to say that yes this has already been thought of.
Notable inventions include:
1. Taking well-known math for trilateration, but putting it inside a phone app and making it ergonomic to use. I vibed it so I can survey a plot quickly using a hand laser ranger and a foldable phone for data entry; apparently, the novelty is in combining the live display of how the measurements resolve to a 2D structure, with error residual, and a list of additional measurements to make, priority-sorted by how much error can be removed by making it. Apparently it's "closing the loop from adjustment back to "what should I physically measure next."" (and yes, it in fact does that, and works well).
2. A shader doing "resolution-independent procedural texturing of a hyperbolic ground plane". Basically the result of me asking "I want something like Hyperbolica to play with on my phone, now now now", followed by "cool but it got ugly tessellation artifacts far from origin". I'm not qualified to judge whether this is in any way not obvious, but Claude is framing this as "specific technical solution to a specific floating-point problem, which is exactly the flavor of thing that survives §101 scrutiny best".
3. There's one I'm afraid to even describe because it has to do with UX of VLMs and that could make it novel enough somebody will patent it. But to hell with it: basically, take an image annotator (draw colored rectangles on an image), allow user to label the rectangles in the interface, and then save, alongside the image, a file mapping colors to labels. This was my idea to avoid having to write stuff like "that red thing is XYZ" and "the purple rectangles are VWX". Pretty obvious UX streamlining that's one of many that apparently no major AI provider thought of yet...
--
1. is a shared invention - me pushing for ergonomics, LLM figuring out the "game loop". 2. was Claude Fable all on its own - I just told it it's ugly and looks like the kind of nonsense you get at the poles of a globe made of triangle strips, and it came up with the shader. 3. was my idea from the start.
Question to patent people: are things like this really patentable? I hope not...
The real problem is patents which are trivial and/broad being granted.
OpenAI, Anthropic, Google, Meta and X are each valued at a trillion or more. Mistral is valued at $20B, 50x less. Even Deepseek is several times bigger than Mistral.
$20B is certainly not poor, but in the world of AI and overinflated valuations, it almost is.
Btw, OpenAI words, not mine https://news.ycombinator.com/item?id=49232778
The USPTO has a strange insistence on granting them even though they aren't legally valid. Therefore, if you have a lot of money, you might file a software patent for one of four reasons:
1) You hope that the US courts might one day change their minds and start enforcing software patents.
2) You hope that you might scare people in to giving you licensing fees even though you know the patent won't be enforced.
3) You want to prevent others from scaring you in to giving them licensing fees.
4) You want to get a patent because it's impressive/looks good on your resume.
otherwise an evil law firm could spam adaptations of your idea to every relevant industry and get rights over your invention without your consent. that would be a complete perversion of what the patent system is supposed to do.
I recently learned [0] that the USPTO makes it money from patents, its not government funded. Not only that but checking patents loses them net money while maintenance fees are the real cash cow.
The whole system is similar to the revenue model of a shitty journal that just publishes whatever research as long as the author pays. Except the office doesn't even need to care about their reputation in granting dubious patents because they have legal backing.
[0] It was a comment on hacker news, that I checked.
It's not the user fee model that's broken, it's the revenue basis. Maybe it would be more effective if they tripled the filing cost and if you successfully obtain the patent (it's rejected by the patent officers), you get half back.
No you don't. That being said, what is the point of a patent if not to enforce the right of exclusivity it grants as to the invention?
It's always something like this, from someone who didn't even realize they weren't actually talking about patents in the first place.
Usually it's to defend against another company using a "blocking patent" by pursuing you for infringing their patent which is built upon the thing you patented, doing the next step which was always obvious to you.
So that you can continue building what you started in the first place, instead of being shut down while someone else takes your idea forward.
Not for exclusivity.
Would be pretty hard to sue someone for practicing a patent that cites that person in prior art who concurrently was inventing the same thing and would say how obvious it ones. You're going to tell me it's not so easy, but you completely made up some really narrow and specific situation anyway...
>Not for exclusivity.
It's only your right to exclude others from it that gives a patent any defense... it's all about exclusivity. That's the fundamental nature of property law.
Company A publishes invention 1, when they know inventions 2, 3, 4 follow naturally in their own roadmap but those things aren't obvious to other people yet. Invention 1 is just a stepping stone in company A's larger vision, and it's not useful by itself, but they can't publish (or patent) everything at once, it takes time to develop them. It's also not permitted to patent 4 inventions in one patent. There are time, expense and R&D factors.
Company B studies invention 1, realises invention 2 follows naturally (as does everyone who looks at invention 1), and company B patents invention 2 before anyone else.
They can do this because inventions 2, 3, 4 are not obvious to persons skilled in the art who haven't seen invention 1. They are only obvious to anyone who works with invention 1.
The standard is not "if you knew about invention 1, would 2 be an obvious next step", it's "would 2 be obvious with general industry knowledge alone".
When that happens company A is stuck, whether they patented invention 1 or just published and/or used it.
With their prior art, they can continue to use invention 1, even if they didn't patent it, but it's not much use by itself. They can't develop it further in the directions they'd had in mind all along. And sometimes company B's patent and company A's prior art will mean company A can continue to use invention 1 but nobody else can.
A defensive strategy is for company A to patent invention 1, even if they hate patents, just so that company B can't use invention 1 when they need it in invention 2. This doesn't stop company B from patenting invention 2 before company A, by the way. It just creates leverage for company A to persuade company B to license invention 2, instead of company A having to abandon its entire roadmap and switch to a different kind of product.
> I's only your right to exclude others from it that gives a patent any defense... it's all about exclusivity
Yes, but not all companies want to exclude. Some even want to share their work openly as they proceed. But even sharing openly and creating prior art as you go is not safe.
The natural line of development they have mapped out can be shut down by others seeing the way ahead and blocking it, which is a strategy, as I said, that I have seen discussed in real patent lawyer meetings.
For those companies, "what is the point of a patent if not to enforce the right of exclusivity it grants" -> the point is not to enforce exclusivity over anyone, because they don't want to. It's to use the granted rights as defensive leverage so they can continue developing, publishing and using their own work.
>Yes, but not all companies want to exclude. Some even want to share their work openly as they proceed. But even sharing openly and creating prior art as you go is not safe.
That's great, patents are publicly available documents. So patentees can feel free to share!
>For those companies, "what is the point of a patent if not to enforce the right of exclusivity it grants" -> the point is not to enforce exclusivity over anyone, because they don't want to. It's to use the granted rights as defensive leverage so they can continue developing, publishing and using their own work.
You're just making some semantic point for a reason I cannot understand. Patents only give the right of exclusion, what a patentee does with that right is up to them... They can trade on it, use it to bargain, give it up, whatever.
Well, trade secrets could also do the trick?
Simple, because the patent is less about protecting an invention than doing business in the US.
The US is more hospitable to software patents than Europe. A US patent also gives Mistral something useful when dealing with companies that already hold large patent portfolios: defensive leverage, cross-licensing, and one more asset on the books to make it attractive to investors.
There's a certain absurdity to it, and the French have a history with absurd. A French open-weights company crosses the Atlantic to put a patent around a loop, not because the code is remarkable, but because the American market makes the loop a valuable asset, business-wise.
You can see the actual claims here: https://ibb.co/k2mVVkFN
You can find the full patent on the USPTO search site
https://ppubs.uspto.gov/pubwebapp/
search: "12670045".did
search: "12670045".did
My reasoning: - That way I will have maximum control over how the LLM can interact with the world (i.e. I can even narrow down which arguments I support/not support) - The monitoring story should be great as well, since every interaction with the outside world happens at one place only - I can add functionality as I go, and can understand how the harness changes and why it does what it does from the ground up
The biggest challenge I foresee is with the LLMs making tool calls and having them react to the results. I also wonder how far I will even get with this :D
Back to topic; "code implemented tool calls" is basically my "oh yeah my harness can do tool calls too" (and everything I need is implemented in one codebase, in... code)
Given I wrote an article describing the same things years before they submitted the patent, could I not invalidate their patent with that?
The big picture here is that overall tool call flow works as normal, but rather than the LLM just generating tool calls, it is instead generating a "code block" that may include multiple tool calls/invocations. The server (between LLM and client) receives this code block and will execute it, but first has to execute any contained tool calls, some which it may do locally, and some via calls back to the client/agent as normal. Once all tools calls and the containing code block have been executed by the server (in a sandbox), then the final result is sent back to the client.
Another step (although governance would need to be super good for this). Would be for this charity to withold access to these patents from companies that decide to abuse stupid patents.
For example, I would publish the idea of a "self-driving car" that can drive without or with minimal human supervision using a computer. I believe this is pretty novel and can be called an invention.
Also I hope this patent is valid only in US and cannot be enforced in China.
The patent office rarely care about it, they register whatever you want, and then lawyers of companies can sue you for using it without permission. Of course you can fight it in court and affirm it's prior art, if you have enough money to pay your lawyer, and do you have more money than OpenAI, Meta, Google? Probably not, so they will just shut you down, that is the sole reason for patents to exist, to preserve monopoly of big multinational companies.
And nowadays that patents are probably written and submitted not even by human but by AI, what stops a big tech with enough money for using a model to generate millions of trivial patent, submit them, and get them approved, for then another AI to search for their usage and sue you for using it?
It requires a yearly payment and can get quite expensive [1]. There are application, examination, issuing and publishing fees for each patent. In contrast, publishing ideas online is free.
[1] https://www.uspto.gov/learning-and-resources/fees-and-paymen...
Publishing ideas online doesn't really matter, a big company can sue you and you have already lost, because you don't have the money to pay lawyers to defend you, even if you are in right.
There is no justice in our system, unfortunately who owns the money just rules. We don't live in a democracy, we live in something that is not much different of medieval feudalism, possibly much worse indeed.
[1] https://martinodersky.substack.com/p/tracked-capabilities-fo...
[2] https://github.com/lampepfl/tacit
[3] https://www.caisconf.org/program/2026/papers/securing-agents...
Also this is a patent application, so it has not been examined yet. The claims at this point will always be over-broad, because that's just how patent prosecution works: the applicant (brazenly) tries to stake out as much "intellectual territory" as possible (because why not!) but the patent authorities will push back to the extent it does not overlap with other existing "territories" i.e. prior art. The applicant then narrows their claims and the back-and-forth continues until the authority determines that the claims are indeed novel and non-obvious enough to be granted. Every alternate back-and-forth costs $$ and time, which incentivizes applicants to make reasonable claims.
A typical way this negotiation often proceeds is that the applicant starts with a couple of the broadest claims they think reasonable and add multiple narrowing claims (or "dependent" claims) that cover narrower and narrower functionalities, essentially a dependency tree of claims.
This is because the examiner must judge each claim on its own merits, so while the broader claims could get knocked out, the narrower claims could survive and can indicate which "territory" is still unclaimed. The applicant can then focus on that (e.g. by making it part of the root independent claim.)
So often what is interesting in patent applicitations are the dependent claims.
For this patent, the current primary claim literally covers the concept of a tool call by an LLM, and as comments point out, this is ridiculously broad and will get knocked out by prior art. But the dependent claims are intriguing.
A quick glance at those indicates that Mistral thinks the more novel or non-obvious aspects are a "stateless resumable sandbox" which works by executing the generated code until a "non-deterministic operation" (like a random number or a date) and on getting the result of that operation replaying all the code that executed up until that point. The results of previous such operations are cached for reuse in subsequent re-executions.
Now I'm not at all knowledgeable about this area, but I did write a very rudimentary tool-calling LLM harness at one point, and offhand I would now have considered this approach -- re-executing the same code multiple times just seems wasteful and unnecessary. But the specification points out there are other considerations at play in more complex situations, so maybe this is narrower and more interesting than what the title and top claim indicates.
You can see the claims here: https://ibb.co/k2mVVkFN
Turns out it was fast-tracked, which is why it was issued in ~2 months. It does claim priority to an earlier filing so the invention date is indeed March 2006. Not a single claim was questioned or rejected, it went straight to grant!
Yeah, unless I'm misreading claim 1, it seems iffy. From a quick glance at the spec, maybe the novel part is that all tool calls are executed in the sandboxed code block, thus keeping the intermediate results from polluting the context. But I'm not sure if that's different from existing systems like LangChain.
Do you know of a jurisdiction that disallows inventions generated with "AI"?
Edit: downvote if you want, but here's another example: "Companies selling artificial intelligence models in Europe should pay a "levy" to support cultural industries, the head of French developer Mistral said on Friday, March 20." https://www.lemonde.fr/en/international/article/2026/03/20/m...
Somehow I doubt that will happen.
It reminds me of those taxes on CD-R disks to support "Art associations" that were proposed before.
Voices like this are useful to spark a debate on AI basically taking content owned by other people and selling it, but no one will agree to implement anything like what is proposed.
They're not just proposed, they are implemented in a lot of countries and France actually has the highest CD tax in the EU (and probably in the world but I haven't checked)
> It reminds me of those taxes on CD-R disks to support "Art associations" that were proposed before.
I find it funny when citizens of a country say things like this. Are you unaware that the Polish government brought in the same tax, 1-3%, on smartphones, laptops, desktop computers and SD cards from November this year? https://tvn24.pl/biznes/tech/nowa-oplata-od-smartfonow-table...
The idea of patents is not to read a feed of the new ones and to discuss them as innovations, it's for the patent holders to use as proof in the future that they had already thought about a thing. The best strategy, in any case, is to publish your own patents, or to document your inventions privately or through your own public channels. Such that in the future IF you are sued for patent infringement, you can prove you didn't arrive at your IP by stealing theirs.
Obligatory: I Am Not A Lawyer
Once worked at a startup with a legendary dumbass who would tell anyone who listened about their patent for URL routing in javascript. Please don't hurt yourself rolling your eyes
[1] … taken out by my employer etc.
So we are just vibe patenting rubbish and rebranding it with AI?
"Patenting rubbish" is a description of pretty much all software patents since forever.
In particular it's the server telling the client to execute arbitrary code in response to a request from the client.
What is the point?