Anthropic is going to watermark all text that comes out of Claude and give companies the tool to detect it, but not the users.
Now they are also blocking specific uses of their models, with a blast radius that also affects people with similar but unrelated use cases.
Add that to all the safeguards they already have added to their models for copyright infringement, security, and other "safety" features.
It does not seem to be a great strategy when faced by fierce competition from cheaper and/or open-weight models from countries like China that do not care about any of these.
If they didn't want us to use other people's ideas without permission then why did they use other people's ideas without permission to make the ideal tool for using other people's ideas with permission?
> If they didn't want us to use other people's ideas without permission then why did they use other people's ideas without permission to make the ideal tool for using other people's ideas with permission?
I don’t think it’s about what Anthropic does or doesn’t want us to use their tools for. They are in the business of selling compute via AI, the more people that use it the better.
This is about stopping future issues (real or perceived) that could lead to legislation curtailing usage of their tools.
Is this really the same cowboys that recklessly broke into others networks and instead of apologising proudly told the world how they did it?
Every 6 months Anthropic publishes a report telling us how closer they have gotten us to extinction as if human life was just a game, yet now they draw the line at computer game modding??
Is this.. The right approach? Who decides what gets to be reverse engineered and what doesn't? I don't have a solution to offer, but I bristle at the idea of one indistry deciding which other industries suffer the negative externalities of their product, while others they decide are spared.
Curious about your setup if you don’t mind elaborating a bit for us? What models are you using and how do you find they compare? The 128gb model costs less than one months salary for a medium level engineer here, they look quite attractive.
(technically it isn't even legal, since changing what people get after they've agreed to the sale is obviously highly illegal)
Not that they've ever actually announced how much a subscription buys you. In fact the token use for a task on the API is also not consistent, so even on the API you don't know. And then there's Sam Altman in interviews, where he says what he wants to do: you submit a prompt, ChatGPT sends you a price and just charges it.
Yes and this is why self-hosting is more important than ever. The issue with LLMs is the cost of hardware if you want something close to the frontier. But there is no excuse not to host your own files, email, media center, etc.
Email is, given the issues with connectivity, something you can justifiably make someone else's problem.
There's no reason not to have your own domain though. I've been using Migadu and the amount of flexibility that gives you is ridiculous. The only downside is the lack of push notifications, but if your email is so important that you can't wait for a refresh then I can't help.
same as self-hosting email/webservers though, you dont need world-class hardware for personal projects. frontier models may be more capable but a smaller model can just as easily work through a project you break down for it
This about the risk of getting sued by parties with deep pockets. Antrophic is close to an IPO so that is definitely not the kind of attention they want.
The money we are talking about.. A lot of these rights holders have more than enough cash on hand to bring a suite and potentially destroy Anthropic/OpenAI with the right case.
Who? Netflix and Disney are the largest media rights holders, and they have market caps of ~$250 billion and ~$190 billion, respectively. OpenAI and Anthropic are both privately held, but currently have valuations of somewhere in the $1-2 trillion range.
While a good rule of thumb for the US court system is that someone with more money can screw over a smaller party. That's only up until a point. The most expensive lawsuit in history by legal fees was (adjusted for inflation) $1.15 billion. That was the 1998 Tobacco masters settlement (involving years of litigation by law firms across 46 states). So any award in the billions makes a case arguably justifiable (even more so when the highest litigation cost cases are often that expensive because the litigation fee includes a percentage of winnings).
Companies like autodesk with a ttm revenue of $7.8 billion would probably have fiscally compelling cases (if not legally compelling) to pursue if reverse engineered tools cut into their revenue. That's without getting into class action cases.
> While a good rule of thumb for the US court system is that someone with more money can screw over a smaller party. That's only up until a point.
Exactly.
Adobe had a FY2025 operating income of 8.7b on 23.77bn revenue. They employ over 33k people..
So Adobe alone has the money for a law suite that defends themselves from an existential threat, but I would imagine a class action.. It is, IMHO, just a matter of time before these companies put it together and take their shot.
> This about the risk of getting sued by parties with deep pockets.
They are also probably intentionally trying to create differentiation between their models and open models to strengthen the regulatory capture case of having the government ban open model use.
I can see your point, but on the other hand it seems like the only alternatives are a blanket ban on reverse engineering or a blanket allowance. I probably don't have to make the case for the value of reverse engineering here, so the question is to what degree it can be abused and if they have the responsibility to stop it. I would argue companies do have an obligation to consider and mitigate abuse of their product, especially frontier technologies. Probably the best outcome will be some kind of balance of judgment by the company.
A blanked allowance would be the answer. Claude is just a tool, after all. Neither Ghidra nor IDA tells me what I'm allowed to decompile and what I'm not. If someone then uses it to create something whose distribution would constitute a copyright infringement, that is the responsibility of the person who did so, not the software manufacturer.
Several AI companies are getting sued for what their products allow users to do, as was Google for what YouTube allowed its users to do.
YouTube is probably an instructive example here: they can refuse to host certain content on their servers, and this is Anthropic refusing to have the AI on their servers perform certain tasks.
Looks like we need to ban paper. Paper can be used to reproduce copyrighted material.
Even without a computer, you could use a pencil and a notebook to write down some song lyrics without express written consent of Universal Music Group!
It's the wrong approach, because the public demand to reverse engineer games (a pretty benign activity, for humans at least not corporations) is massive and a new subset of hackers who will release public jailbreaks that could be used for hacking which is more of a grey area
We're already knocking on the door of just clean-room recreating any game or experience in one shot. In such an environment, who really cares about reverse engineering old games?
I mean, from game company perspective, will they really care about their old games getting reverse engineered when their next big title can be clean-room recreated by anyone with the skill to write a sentence? Maybe creating a game similar to the next big release will be something that the AIs refuse to do next? Interesting times.
As an aspiring game developer myself, I guess reducing the accessibility of old games benefits me (but I don't like it)--and hey, if you can't get that old game to run, maybe we can vibe code for a weekend and make a new game that captures the same feeling.
> Who decides what gets to be reverse engineered and what doesn't?
The service provider decides the terms and conditions, as long as they comply with the law. For example, Google decides whether pirated movies can be hosted on its cloud service, because it provides the cloud service. It's pretty clear.
In a pure capitalist system, competitors would emerge if the repressed service were genuinely in high demand and lawful. If it was not, then well, that's just how we regulate the world but this cannot be considered a case of one industry deciding the fate of another.
Honestly the collateral worries me more than the game policy itself. The Tom's Hardware piece mentions users saying generic reverse-engineering tasks get blocked too, and a classifier looking at disassembly can't easily tell a commercial game from your own firmware, a binary you're patching for interop, or a CTF challenge. If a filter is going to draw that line, I'd rather it were documented than discovered one refusal at a time.
Game REs & crackers are probably the most persistent personality types you can find in software world. They are also one of the most technically sophisticated software engineers around. I expect we will see a new AI system pop-up soon, made by RE, specifically trained to RE games/software, thanks to this ban. Pandora box is open.
They really didn’t though. They put it in a client side modification to the system prompt. You can just rip it out with a proxy or a small local modification to your client, which Claude will be happy to do for you. They took a token step so that people would write these headlines in the hope it would stave off more negative attention in the video game world without actually doing anything meaningful.
IIRC Claude has some kind of mechanism where even though the system prompt is sent from the client, if it’s modified it can trip tamper checks and result in a ban.
If they tried to stop terrorists from making biological weapons in the same manner, the headlines would call that useless and ineffective because you're trying to stop already bad actors with something ineffective and trivially bypassed. This is no different, and shouldn't be treated any differently.
I wanted to reverse engineer a win-xp era game with boring music and make it play my own selections--inherently a fair-use activity. The game was somewhat obfuscated, so while I've used ghidra successfully at work, I wasn't able to work out how the game's system design even after a saturday of poking at it.
Claude has already refused to perform incredibly not-security-relevant tasks, so I'm guessing it will simply not help with this now either. Today, I am, for the first time, looking into my own LLM hardware.
I think that, excepting part of the HN crowd and linkedin, literally the only positive comments I've heard about AI have been related to game mods.
I'm saying this as a completely orthogonal conversation to regularisation, ethics, etc. It's just amazing to see how little value these companies place in the general public's perception of them.
Can you think of any other companies selling to the mass marker that care so little about it?
I think that they think, not necessarily incorrectly, that their tool is applicable across any and all domains, so restricting a few of these domains is small fry in the overall big picture.
What they may be missing is that the domains in which it it's being used to great success is the pipeline to the expansion of is use into other domains.
Having said that, the potential legal issues of being a reverse engineering tool is a strong legal motivator. But I also doubt that any legal issues as a result of reverse engineering could be fatal to either Anthropic or AI/LLM usage; the tool, being general purpose, will allow for the legal responsibility to be put upon the person using it, especially if it's decided that Anthropic and OpenAI are protected national security assets.
On the contrary, this reads to me as a response to the "we're taking down our game demos because we can't risk AI decomp-cloning attempts" and "oh you released a promising demo for your indie game? Pay me $200 not to make a slop clone of it and release it before you" incidents that have gathered attention outside HN, and it's the average HN commenter that's misinterpreting public perception on this specific topic because the hobbyist classic-game recomps dominate this feed.
The things that normal people will want to use AI for (game mods) will be made illegal or a ToS violation until the only thing allowed is to consume veiled advertisements.
Behind closed doors the elite will use the tech for (biological) weapons and reverse engineering though. Military applications, OK but don’t you dare put Sonic in Super Mario.
I am so glad I was already able to reconstruct true-to-the-originals C++ versions of Colossal Cave Adventure, Zork (mainframe version), Mystery Mansion and Warp.
Those are the ones that matter. :)
I don't see any company's blockade on this kind of activity having much impact. Too many models. Too much fun.
i ported the whole DungeonMaster 2 to .net. 90% with the first few 20 prompts, then quite some many to make it perfect. Just from the executables from those DosBox sites. Amazing.
Not from my experience. Missing 10% were caused by medium effort and that agent was not able to do play tests. So it is just testing and issuing additional prompts. Nothing hard. No thinking required.
Very sad. I'm about 50% in taking the old Red Alert 2 and get to something that would've resembled the original source code.
Given the rumors that the source code has been lost (and the fact that they open sourced the games before Tiberian Sun & Red Alert 2, and the ones after those), I felt like I was using the tokens for 'good'.
This is a good canary of what people were discussing in the past. Left to its inevitable outcome, centralizing AI development of software in a central player will inevitably generate over control and abuse. Its fortunate that openai or other party did not get to monopolize the Ai coding game and we have a way out in commoditization and open models.
Yeah, if they cripple the ability to mod video games, that'll result in a lot of users learning how to use alternative open weight models without all the refusals.
Anthropic really needs to be considering the fact that a lot of those people will be working in IT--and when their bosses start complaining about the checks they're writing to Anthropic, they'll have solid experience with alternative solutions to suggest.
They should be careful not to kneecap their product which would damage their valuation and number of subscribers in favor of trying to stop the inevitable.
I am busy working on an embarrassingly ambitious game reverse engineering project. (To the point where admitting what it is feels like telling someone you’re working on a perpetual motion machine).
It’s a new tool to reverse engineer games in a specific context, and I’ve been using anthropic models to work on it.
I haven’t seen any increase in safeguards triggering or any negative effects in how quickly agents are able to make progress in this space.
To give an idea of scope it’s about 5000 commits, 500k lines of rust code, 100 Ghidra patches of work so far. Made in around 6 weeks worth of 8-15 agents working continuously.
The most visible productivity gain I've seen from AI is in this kind of cool hoby RE stuff that just wasn't worth the time investment/resources before.
They probably have reverse engineered the stuff for themselves and now closing the door behind them. That type of thing is quite common. For
example MSFT copied some other operating system itself but theirs was holy. OpenAi used google's open source ai tools before openAi.
Fable of The Scorpion and the Frog should always apply .
they haven't seen anything yet, this is just the beginning. reverse-engineering is here to stay and there's nothing anyone can do about it. they brought the idea of AI to make us jobless, now the same AI is reciprocating the same energy and they don't like it?
"Anthropic's Claude AI has quietly started restricting the kinds of work it will do, including modding, decompiling, or reverse-engineering video game code."
Uh, we're not allowed to use Claude to even mod games that fully support mods?
Based on how every time I ask it a virology question about SARS-CoV-2 and I get a refusal, I'm quite worried about how far-reaching this will be.
Oh no little piglet, you’re not supposed to be eating at the dinner table with us. Your place is at the trough. Make sure you eat all your slop to grow up big and fat. Do that and I’m sure you’ll find yourself on the dinner table in no time.
It's not a surprise after so many IP violations recently that big guys will start blocking this stuff. But people who wants to reverse-engineer will just use Chinese or local models for that ¯\_(ツ)_/¯
I don't think it's going to help Anthropic / OpenAI with that fight, but this is something that have to do to make their investors happy
Why does anthropic feel the need to babysit every one of their users? its so disgusting. Look - i get bioterrorism, weapons, etc and why you may need guardrails on those, but RE' games? come on.
I wonder when someone will offer a truly uncensored model subscription. For some reason, it seems that nobody's been brave enough to find a country that wouldn't take them down for offering that service.
This is just prompt noise that any RE will be able to easily coax claude to ignore. It exists to get the big loserly game companies like Nintendo and Sony off their back.
> Like Claude Mythos Preview, GLM-5.3 has strong capabilities for autonomously building end-to-end cyber exploits. But GLM-5.3 is unlike other frontier models in that it has been released without meaningful safeguards to limit misuse. https://www.anthropic.com/research/glm-5-3-and-the-spread-of...
First, they came for the people who wanted to make kinky art, and I did not speak out, for I am a Sports Illustrated Swimsuit Edition kinda dude...
Then, they came for the people who wanted to reverse-engineer video games, and I did not speak out, for I find no amusement in seeing Mario take down Elden Ring bosses...
Finally, they banned me for a politically incorrect/inconvenient prompt (like asking for help defending Luigi in a debate or something), and there was no one left to speak out for me.
Why was there no one left to speak for you. Did the AI apocalypse actually happen and it is just you and a rack running an instance of "Claude" left on earth?
The cost of blocking videogame reverse engineering is trivial. They added three lines to the system prompt to exclude it and probably made a small update to their backend training classifiers. They give up the videogame reverse engineering market, which is a tiny future revenue stream and not worth much financially.
In exchange they avoid hundreds of millions, maybe billions, in future copyright lawsuits.
Providing this service has negative expected value. It's better to cede the market to someone else.
> In exchange they avoid hundreds of millions, maybe billions, in future copyright lawsuits.
Will the lawsuits not come from adobe? From big SaaS companies? From every other major piece of software any Joe will be able to duplicate with a single prompt in a few hours?
This admission is showing that Anthropic believes they might get in hot water for reverse engineering video games.
Surely it applies to everything else.
They are absolutely going to be getting sued by major software IP and copyright holders.
Adobe and other companies have been holding back because they are also using frontier models to keep pace. Suing Anthropic means getting cut off. But now that alternatives are proving themselves..
This, to me, is why Anthropic is rushing their IPO.
I simply can't imagine a company of Adobe's size (employing over 33k people and making 24bn annual revenue) is going to sit on the sidelines while this plays out.
And they are just one company. All faced with a technology that can simply copy all their hard work and investment in product development.
Video game reverse engineering is the biggest and most visible problem in the present moment, it's widely discussed in the media, and there are video game companies known to be very litigious against people who reverse engineering their products (ex: Nintendo). It makes sense to block the biggest problem first.
If future lawsuits come from other industries, then they'll likely add a blanket ban on other forms of reverse engineering.
> They give up the videogame reverse engineering market, which is a tiny future revenue stream and not worth much financially
Btw video games are by far the biggest entertainment industry in history.
If the way video games of the future are made is with AI, the logic might be reversed; giving up some lawsuits now, while missing out on large entertainment revenue in the long term.
But that's not at all what Anthropic is doing. Anthropic is not blocking people from creating video games with AI. Using Claude to create new video games is completely valid and unrestricted; they just don't want people to use it for reverse engineering existing ones, as that opens them up to massive legal liability.
I agree this is about money and they don't want to be legally liable.
But Minecraft is reverse engineered Infiniminer. Fortnite is reverse engineered PubG. LoL is reverse engineered DotA. This is literally how the industry works.
The next best thing is gonna be some degree of RE of existing stuff -- that's clearly what people want. Anthropic is just trying to placate the old guard, denying users what they want, and ultimately will be on the wrong side of history on this.
That's why I'm saying their move seems extremely short-sighted. They're locked in on the IPO at the cost of the future.
You are confusing game design inspiration with code/asset reverse engineering, which are entirely different things.
Taking inspiration from other games, i.e. implementing similar gameplay loops or mechanics from other games, is completely normal within the industry. Unless copyrighted, it's not illegal. Minecraft taking from Infiniminer, or LoL taking from DotA is industry practice. Claude doesn't prohibit this.
Reverse engineering, i.e. decompiling game binaries to extract source code and copyrighted assets, then generating derivative games using said materials, is grounds for a massive copyright lawsuit. Anthropic doesn't want a part of this. That's why Anthropic is preemptively prohibiting it.
Anthropic is going to watermark all text that comes out of Claude and give companies the tool to detect it, but not the users.
Now they are also blocking specific uses of their models, with a blast radius that also affects people with similar but unrelated use cases.
Add that to all the safeguards they already have added to their models for copyright infringement, security, and other "safety" features.
It does not seem to be a great strategy when faced by fierce competition from cheaper and/or open-weight models from countries like China that do not care about any of these.
If they didn't want us to use other people's ideas without permission then why did they use other people's ideas without permission to make the ideal tool for using other people's ideas with permission?
> If they didn't want us to use other people's ideas without permission then why did they use other people's ideas without permission to make the ideal tool for using other people's ideas with permission?
Because they're all about what's best for them.
They want them to steal everything, they just don't want us to.
It seems all too common these days for tech startups to try to wall off a common good and try to sell access to what used to be free
These days? This is how spotify was found. Its same playbook. Steal everything and then do regulatory capture.
I don’t think it’s about what Anthropic does or doesn’t want us to use their tools for. They are in the business of selling compute via AI, the more people that use it the better.
This is about stopping future issues (real or perceived) that could lead to legislation curtailing usage of their tools.
These guys are no fun at all.
Is this really the same cowboys that recklessly broke into others networks and instead of apologising proudly told the world how they did it?
Every 6 months Anthropic publishes a report telling us how closer they have gotten us to extinction as if human life was just a game, yet now they draw the line at computer game modding??
And stole every bit of Intellectual Property they could get their hands on legally or otherwise.
For my friends, everything, for my enemies, the law.
Is this.. The right approach? Who decides what gets to be reverse engineered and what doesn't? I don't have a solution to offer, but I bristle at the idea of one indistry deciding which other industries suffer the negative externalities of their product, while others they decide are spared.
> Who decides what gets to be reverse engineered and what doesn't?
whoever owns the hardware
currently running 2x DGX Sparks and never thought to use a commercial model again
Curious about your setup if you don’t mind elaborating a bit for us? What models are you using and how do you find they compare? The 128gb model costs less than one months salary for a medium level engineer here, they look quite attractive.
A month and a half salary after tax here. That is still particularly expensive, and besides its sinking money into a steeply depreciating asset.
You mean appreciating? They have doubled in price since release...
That’s about five months of wages up here in rural nowhere NY doing IT, for what it’s worth.
Some of us are just out of the AI race completely.
i've been running DeepSeekV4 flash but looking at the other models being suggested around, best part is switching to test has 0 cost :)
i felt like commercials LLMs were "good enough" for my needs about 2-3 generations ago, so this clicked into my work/life needs perfectly
i got them right before they went up in price, but honestly, for the ability to know that i wont have my thinking "questioned"... priceless
Best model on 2xSpark currently is GLM 5.3 flash ... but really does require 2x Spark. 25 to 60 tok/s with MTP.
2x Spark? As in two DGX Spark machines? That is obscenely expensive.
Is it? How much would it cost to hire a person to do the same work two of these can do?
Maybe fair to compare to using AI subscription ~20$/m (not Anthropic neither Openai)
Especially now with OpenAI constantly changing (and always reducing) what the subscription can do ... At least it will keep working.
Right now I need 4 subscriptions to do what 1 used to do. And they're tightening it further, without announcing:
https://news.ycombinator.com/item?id=50320588
(technically it isn't even legal, since changing what people get after they've agreed to the sale is obviously highly illegal)
Not that they've ever actually announced how much a subscription buys you. In fact the token use for a task on the API is also not consistent, so even on the API you don't know. And then there's Sam Altman in interviews, where he says what he wants to do: you submit a prompt, ChatGPT sends you a price and just charges it.
> whoever owns the hardware
Yes and this is why self-hosting is more important than ever. The issue with LLMs is the cost of hardware if you want something close to the frontier. But there is no excuse not to host your own files, email, media center, etc.
Email is, given the issues with connectivity, something you can justifiably make someone else's problem.
There's no reason not to have your own domain though. I've been using Migadu and the amount of flexibility that gives you is ridiculous. The only downside is the lack of push notifications, but if your email is so important that you can't wait for a refresh then I can't help.
same as self-hosting email/webservers though, you dont need world-class hardware for personal projects. frontier models may be more capable but a smaller model can just as easily work through a project you break down for it
This about the risk of getting sued by parties with deep pockets. Antrophic is close to an IPO so that is definitely not the kind of attention they want.
There are no content owners that have close to as deep of pockets as the big AI companies
The money we are talking about.. A lot of these rights holders have more than enough cash on hand to bring a suite and potentially destroy Anthropic/OpenAI with the right case.
Who? Netflix and Disney are the largest media rights holders, and they have market caps of ~$250 billion and ~$190 billion, respectively. OpenAI and Anthropic are both privately held, but currently have valuations of somewhere in the $1-2 trillion range.
That is not a fair fight.
While a good rule of thumb for the US court system is that someone with more money can screw over a smaller party. That's only up until a point. The most expensive lawsuit in history by legal fees was (adjusted for inflation) $1.15 billion. That was the 1998 Tobacco masters settlement (involving years of litigation by law firms across 46 states). So any award in the billions makes a case arguably justifiable (even more so when the highest litigation cost cases are often that expensive because the litigation fee includes a percentage of winnings).
Companies like autodesk with a ttm revenue of $7.8 billion would probably have fiscally compelling cases (if not legally compelling) to pursue if reverse engineered tools cut into their revenue. That's without getting into class action cases.
> While a good rule of thumb for the US court system is that someone with more money can screw over a smaller party. That's only up until a point.
Exactly.
Adobe had a FY2025 operating income of 8.7b on 23.77bn revenue. They employ over 33k people..
So Adobe alone has the money for a law suite that defends themselves from an existential threat, but I would imagine a class action.. It is, IMHO, just a matter of time before these companies put it together and take their shot.
> This about the risk of getting sued by parties with deep pockets.
They are also probably intentionally trying to create differentiation between their models and open models to strengthen the regulatory capture case of having the government ban open model use.
Can't put the cat back in the tube
I can see your point, but on the other hand it seems like the only alternatives are a blanket ban on reverse engineering or a blanket allowance. I probably don't have to make the case for the value of reverse engineering here, so the question is to what degree it can be abused and if they have the responsibility to stop it. I would argue companies do have an obligation to consider and mitigate abuse of their product, especially frontier technologies. Probably the best outcome will be some kind of balance of judgment by the company.
A blanked allowance would be the answer. Claude is just a tool, after all. Neither Ghidra nor IDA tells me what I'm allowed to decompile and what I'm not. If someone then uses it to create something whose distribution would constitute a copyright infringement, that is the responsibility of the person who did so, not the software manufacturer.
Several AI companies are getting sued for what their products allow users to do, as was Google for what YouTube allowed its users to do.
YouTube is probably an instructive example here: they can refuse to host certain content on their servers, and this is Anthropic refusing to have the AI on their servers perform certain tasks.
> Claude is just a tool
I dont think Anthropic will agree with that statement.
Looks like we need to ban paper. Paper can be used to reproduce copyrighted material.
Even without a computer, you could use a pencil and a notebook to write down some song lyrics without express written consent of Universal Music Group!
It's the wrong approach, because the public demand to reverse engineer games (a pretty benign activity, for humans at least not corporations) is massive and a new subset of hackers who will release public jailbreaks that could be used for hacking which is more of a grey area
Ultimately, the person offering the service gets to decide what the service is.
If I want to make a knitting forum, im not going to feel bad about deleting posts about reverse engineering video games.
We're already knocking on the door of just clean-room recreating any game or experience in one shot. In such an environment, who really cares about reverse engineering old games?
I mean, from game company perspective, will they really care about their old games getting reverse engineered when their next big title can be clean-room recreated by anyone with the skill to write a sentence? Maybe creating a game similar to the next big release will be something that the AIs refuse to do next? Interesting times.
As an aspiring game developer myself, I guess reducing the accessibility of old games benefits me (but I don't like it)--and hey, if you can't get that old game to run, maybe we can vibe code for a weekend and make a new game that captures the same feeling.
> who really cares about reverse engineering old games?
Tetris copyright for example is still heavily policed.
I think however if games continue to be killed there should be a legal route towards the community taking ownership in whatever way is necessary.
Interesting example because there is no lack of Tetris-like games.
"Clean room" is doing a lot of heavy..
Anyway, 99.99% of "clean room" vibe coded projects are 100% NOT putting in place the proper policies and controls.
It's the wild west out there right now.
I think from a legal perspective modding is technically a violation of the ToS.
But ofcourse the videogame industry has always turned a blind eye to it provided you don't make money from it.
> Who decides what gets to be reverse engineered and what doesn't?
Laws, ideally.
> Who decides what gets to be reverse engineered and what doesn't?
The service provider decides the terms and conditions, as long as they comply with the law. For example, Google decides whether pirated movies can be hosted on its cloud service, because it provides the cloud service. It's pretty clear.
In a pure capitalist system, competitors would emerge if the repressed service were genuinely in high demand and lawful. If it was not, then well, that's just how we regulate the world but this cannot be considered a case of one industry deciding the fate of another.
Honestly the collateral worries me more than the game policy itself. The Tom's Hardware piece mentions users saying generic reverse-engineering tasks get blocked too, and a classifier looking at disassembly can't easily tell a commercial game from your own firmware, a binary you're patching for interop, or a CTF challenge. If a filter is going to draw that line, I'd rather it were documented than discovered one refusal at a time.
Game REs & crackers are probably the most persistent personality types you can find in software world. They are also one of the most technically sophisticated software engineers around. I expect we will see a new AI system pop-up soon, made by RE, specifically trained to RE games/software, thanks to this ban. Pandora box is open.
They really didn’t though. They put it in a client side modification to the system prompt. You can just rip it out with a proxy or a small local modification to your client, which Claude will be happy to do for you. They took a token step so that people would write these headlines in the hope it would stave off more negative attention in the video game world without actually doing anything meaningful.
IIRC Claude has some kind of mechanism where even though the system prompt is sent from the client, if it’s modified it can trip tamper checks and result in a ban.
Sounds like an illegal DMCA circumvention device. Handcuffs! (IANAL)
Can you ask Claude to help with the modification of the local client?
You’re literally describing how they did set up a block, while also saying they didn’t
If they tried to stop terrorists from making biological weapons in the same manner, the headlines would call that useless and ineffective because you're trying to stop already bad actors with something ineffective and trivially bypassed. This is no different, and shouldn't be treated any differently.
Sure, that’s still a block. A client side block is obviously a block
They're literally describing how it's not actually a block just a minor inconvenience
A block would be a bit harder to circumvent
It’s a client side block… which is of course a block
I wanted to reverse engineer a win-xp era game with boring music and make it play my own selections--inherently a fair-use activity. The game was somewhat obfuscated, so while I've used ghidra successfully at work, I wasn't able to work out how the game's system design even after a saturday of poking at it.
Claude has already refused to perform incredibly not-security-relevant tasks, so I'm guessing it will simply not help with this now either. Today, I am, for the first time, looking into my own LLM hardware.
FWIW, Codex is a lot more permissive.
GLM 5.3 via Mistral Subscription would be my recommendation
I think that, excepting part of the HN crowd and linkedin, literally the only positive comments I've heard about AI have been related to game mods.
I'm saying this as a completely orthogonal conversation to regularisation, ethics, etc. It's just amazing to see how little value these companies place in the general public's perception of them.
Can you think of any other companies selling to the mass marker that care so little about it?
I think that they think, not necessarily incorrectly, that their tool is applicable across any and all domains, so restricting a few of these domains is small fry in the overall big picture.
What they may be missing is that the domains in which it it's being used to great success is the pipeline to the expansion of is use into other domains.
Having said that, the potential legal issues of being a reverse engineering tool is a strong legal motivator. But I also doubt that any legal issues as a result of reverse engineering could be fatal to either Anthropic or AI/LLM usage; the tool, being general purpose, will allow for the legal responsibility to be put upon the person using it, especially if it's decided that Anthropic and OpenAI are protected national security assets.
most of their revenue come from a few companies, they couldn't care less about the mass market's opinions of them
On the contrary, this reads to me as a response to the "we're taking down our game demos because we can't risk AI decomp-cloning attempts" and "oh you released a promising demo for your indie game? Pay me $200 not to make a slop clone of it and release it before you" incidents that have gathered attention outside HN, and it's the average HN commenter that's misinterpreting public perception on this specific topic because the hobbyist classic-game recomps dominate this feed.
The things that normal people will want to use AI for (game mods) will be made illegal or a ToS violation until the only thing allowed is to consume veiled advertisements.
Behind closed doors the elite will use the tech for (biological) weapons and reverse engineering though. Military applications, OK but don’t you dare put Sonic in Super Mario.
I am so glad I was already able to reconstruct true-to-the-originals C++ versions of Colossal Cave Adventure, Zork (mainframe version), Mystery Mansion and Warp.
Those are the ones that matter. :)
I don't see any company's blockade on this kind of activity having much impact. Too many models. Too much fun.
i ported the whole DungeonMaster 2 to .net. 90% with the first few 20 prompts, then quite some many to make it perfect. Just from the executables from those DosBox sites. Amazing.
The first 90% is always the easy part. The second 90% is the hard part.
Not from my experience. Missing 10% were caused by medium effort and that agent was not able to do play tests. So it is just testing and issuing additional prompts. Nothing hard. No thinking required.
Small discussion (20 points, 9 comments) https://news.ycombinator.com/item?id=50316635
Very sad. I'm about 50% in taking the old Red Alert 2 and get to something that would've resembled the original source code.
Given the rumors that the source code has been lost (and the fact that they open sourced the games before Tiberian Sun & Red Alert 2, and the ones after those), I felt like I was using the tokens for 'good'.
We'll see when it actually stops.
Might want to build on other people's work to save yourself some time.
Decompilation of RA2
https://github.com/r-melvin/gagarin
Static recompilation of RA2
https://github.com/sp00nznet/redalert2-recomp
We are vibecoder, we push NIH to the max.
I wonder if you can sort of decompile the assembly into basic blocks first, and then only run LLMs on batches of those basic blocks.
How would they know who's the original author or what the rights situation is?
This is a good canary of what people were discussing in the past. Left to its inevitable outcome, centralizing AI development of software in a central player will inevitably generate over control and abuse. Its fortunate that openai or other party did not get to monopolize the Ai coding game and we have a way out in commoditization and open models.
Yeah, if they cripple the ability to mod video games, that'll result in a lot of users learning how to use alternative open weight models without all the refusals.
Anthropic really needs to be considering the fact that a lot of those people will be working in IT--and when their bosses start complaining about the checks they're writing to Anthropic, they'll have solid experience with alternative solutions to suggest.
Are they trying to push more users towards Chinese models?
I don't get it. What does this solve? Wasn't this tool just a skill file with a bunch of scripts? Can't any above-average harness and model drive it?
They should be careful not to kneecap their product which would damage their valuation and number of subscribers in favor of trying to stop the inevitable.
I am busy working on an embarrassingly ambitious game reverse engineering project. (To the point where admitting what it is feels like telling someone you’re working on a perpetual motion machine).
It’s a new tool to reverse engineer games in a specific context, and I’ve been using anthropic models to work on it.
I haven’t seen any increase in safeguards triggering or any negative effects in how quickly agents are able to make progress in this space.
To give an idea of scope it’s about 5000 commits, 500k lines of rust code, 100 Ghidra patches of work so far. Made in around 6 weeks worth of 8-15 agents working continuously.
Out of curiosity what models do you use? How much would a project like that cost?
The most visible productivity gain I've seen from AI is in this kind of cool hoby RE stuff that just wasn't worth the time investment/resources before.
They probably have reverse engineered the stuff for themselves and now closing the door behind them. That type of thing is quite common. For example MSFT copied some other operating system itself but theirs was holy. OpenAi used google's open source ai tools before openAi.
Fable of The Scorpion and the Frog should always apply .
Unfathomably lame
So they own a bio lab but we cant even reverse engineer? Touché, Anthropic
they haven't seen anything yet, this is just the beginning. reverse-engineering is here to stay and there's nothing anyone can do about it. they brought the idea of AI to make us jobless, now the same AI is reciprocating the same energy and they don't like it?
"Anthropic's Claude AI has quietly started restricting the kinds of work it will do, including modding, decompiling, or reverse-engineering video game code."
Uh, we're not allowed to use Claude to even mod games that fully support mods?
Based on how every time I ask it a virology question about SARS-CoV-2 and I get a refusal, I'm quite worried about how far-reaching this will be.
Oh no little piglet, you’re not supposed to be eating at the dinner table with us. Your place is at the trough. Make sure you eat all your slop to grow up big and fat. Do that and I’m sure you’ll find yourself on the dinner table in no time.
This is why we need open models.
The new copyright protects vibes
It's not a surprise after so many IP violations recently that big guys will start blocking this stuff. But people who wants to reverse-engineer will just use Chinese or local models for that ¯\_(ツ)_/¯ I don't think it's going to help Anthropic / OpenAI with that fight, but this is something that have to do to make their investors happy
anthropic can go suck it
Why does anthropic feel the need to babysit every one of their users? its so disgusting. Look - i get bioterrorism, weapons, etc and why you may need guardrails on those, but RE' games? come on.
[dupe] https://news.ycombinator.com/item?id=50304358
https://news.ycombinator.com/item?id=50315987
Its not reverse engineering video games. Its being trained on them
I wonder when someone will offer a truly uncensored model subscription. For some reason, it seems that nobody's been brave enough to find a country that wouldn't take them down for offering that service.
This is just prompt noise that any RE will be able to easily coax claude to ignore. It exists to get the big loserly game companies like Nintendo and Sony off their back.
Also from Anthropic:
> Like Claude Mythos Preview, GLM-5.3 has strong capabilities for autonomously building end-to-end cyber exploits. But GLM-5.3 is unlike other frontier models in that it has been released without meaningful safeguards to limit misuse. https://www.anthropic.com/research/glm-5-3-and-the-spread-of...
Just sayin'...
Thanks for the recommendation, Anthropic!
The irony is...
More paternalistic nonsense
First, they came for the people who wanted to make kinky art, and I did not speak out, for I am a Sports Illustrated Swimsuit Edition kinda dude...
Then, they came for the people who wanted to reverse-engineer video games, and I did not speak out, for I find no amusement in seeing Mario take down Elden Ring bosses...
Finally, they banned me for a politically incorrect/inconvenient prompt (like asking for help defending Luigi in a debate or something), and there was no one left to speak out for me.
Why was there no one left to speak for you. Did the AI apocalypse actually happen and it is just you and a rack running an instance of "Claude" left on earth?
No "fair use" claims this time? /s
A very rational choice by Anthropic.
The cost of blocking videogame reverse engineering is trivial. They added three lines to the system prompt to exclude it and probably made a small update to their backend training classifiers. They give up the videogame reverse engineering market, which is a tiny future revenue stream and not worth much financially. In exchange they avoid hundreds of millions, maybe billions, in future copyright lawsuits.
Providing this service has negative expected value. It's better to cede the market to someone else.
>copyright lawsuits
Absolutely the thing they care about. https://stanforddaily.com/2026/10/27/professors-books-pirate...
Old news and irrelevant. That incident happened in 2021-2022 when Anthropic was a tiny startup.
Now they're a big company and have actual lawyers. They won't make the same mistake again.
> In exchange they avoid hundreds of millions, maybe billions, in future copyright lawsuits.
Will the lawsuits not come from adobe? From big SaaS companies? From every other major piece of software any Joe will be able to duplicate with a single prompt in a few hours?
This admission is showing that Anthropic believes they might get in hot water for reverse engineering video games. Surely it applies to everything else.
They are absolutely going to be getting sued by major software IP and copyright holders.
Adobe and other companies have been holding back because they are also using frontier models to keep pace. Suing Anthropic means getting cut off. But now that alternatives are proving themselves..
This, to me, is why Anthropic is rushing their IPO.
I simply can't imagine a company of Adobe's size (employing over 33k people and making 24bn annual revenue) is going to sit on the sidelines while this plays out.
And they are just one company. All faced with a technology that can simply copy all their hard work and investment in product development.
Video game reverse engineering is the biggest and most visible problem in the present moment, it's widely discussed in the media, and there are video game companies known to be very litigious against people who reverse engineering their products (ex: Nintendo). It makes sense to block the biggest problem first.
If future lawsuits come from other industries, then they'll likely add a blanket ban on other forms of reverse engineering.
> They give up the videogame reverse engineering market, which is a tiny future revenue stream and not worth much financially
Btw video games are by far the biggest entertainment industry in history.
If the way video games of the future are made is with AI, the logic might be reversed; giving up some lawsuits now, while missing out on large entertainment revenue in the long term.
But that's not at all what Anthropic is doing. Anthropic is not blocking people from creating video games with AI. Using Claude to create new video games is completely valid and unrestricted; they just don't want people to use it for reverse engineering existing ones, as that opens them up to massive legal liability.
I agree this is about money and they don't want to be legally liable.
But Minecraft is reverse engineered Infiniminer. Fortnite is reverse engineered PubG. LoL is reverse engineered DotA. This is literally how the industry works.
The next best thing is gonna be some degree of RE of existing stuff -- that's clearly what people want. Anthropic is just trying to placate the old guard, denying users what they want, and ultimately will be on the wrong side of history on this.
That's why I'm saying their move seems extremely short-sighted. They're locked in on the IPO at the cost of the future.
did the agents convince you of an alternative definition of reverse engineering?
You are confusing game design inspiration with code/asset reverse engineering, which are entirely different things.
Taking inspiration from other games, i.e. implementing similar gameplay loops or mechanics from other games, is completely normal within the industry. Unless copyrighted, it's not illegal. Minecraft taking from Infiniminer, or LoL taking from DotA is industry practice. Claude doesn't prohibit this.
Reverse engineering, i.e. decompiling game binaries to extract source code and copyrighted assets, then generating derivative games using said materials, is grounds for a massive copyright lawsuit. Anthropic doesn't want a part of this. That's why Anthropic is preemptively prohibiting it.