This is why we need open AI (not OpenAI). Some duopoly deciding what goes on a whim is not acceptable
I can sorta see them suppressing instructions for say bioweapons (that seems fair to me), but this demonstrates that they'll bend for corporate interests.
> preemptively protecting themselves from potential lawsuits ... for offering the means to reverse-engineer their software.
I think the legal concern isn't so much the act of reverse engineering itself which is mostly legal (except RE specifically to remove anti-copying protection under DMCA), it's that damages from contributory infringement are theoretically unbounded because it's up to $150k per work and $2,500 per download. When the client is worth billions, corporate counsel is going to default maximally conservative, which is extra tragic because secondary infringement almost certainly doesn't apply to AI in this context. Meanwhile, overreacting to this extremely unlikely threat IS causing significant disruption to innocent activities.
Under copyright law secondary infringement doesn't apply to "general purpose technology". It requires "encouraging or knowing in advance it'll be used for infringement," set in the early 80s in the Sony BetaMax case. Under DMCA it requires being "primary designed to circumvent access controls," set in the late 90s by Hollywood lobbyists targeting DeCSS for DVDs. Lawyers today are only defaulting to prohibition because there hasn't yet been a case ruling AI is a general purpose technology. Yet it's obviously the most 'general purpose' tech since, maybe.... the wheel?
OAI / ANT management needs to revisit the secondary copyright infringement prohibition because it's hurting not only their product and brand but also the entire AI value prop. A helpful intelligent agent that randomly goes AWOL for irrationally mistaken reasons is a serious product issue. While mistaken concerns around bioweapons and IT security are understandable, dropping secondary copyright infringement prohibitions seems like an easy win.
The real liability to OAI/Ant is much smaller than it appears, while the long-term downsides are larger than they look. Pragmatically, copyright trolls are just a nuisance, the real threat is someone with deep pockets like Nintendo/Sony, but those guys don't want to sue anyone who might go to court. The last thing they want is a clear ruling that AI is general purpose which would forever remove their vague threat.
They never expected their models could be used for reverse engineering.That's the truth.
Now, they're panicking. And this is why the other dude, OpenAI, came out to say that people shouldn't use RAM locally, because they can't control the software anymore.
Open weight models, ablated models are going to become the norm, hopefully, locally run.
So instead, they're saying, look, hardware should become hard to reach for all of us, for the consumers.It's not because we're competing for resources for the same silicon or chips as they are.The ship has sailed thankfully.
One of the new threat models of local LLMs, especially abliterated ones, is hidden instructions or activation words that obscured trigger tool calls which exfil data.
For people touting about how their software can destroy the world, they aren't awfully creative on how it can achieve that. Hacking breaches of IP was probably most cynicists' top 3 questions asked when presented with the idea of "boosting productivity". If you make it "10x" faster to write code, you make it 10x faster for a hacker to hack. Naive cryptographic principles are less about an ironclad security and more about encryption that takes too much time to brute force. It all leads to one thing.
I'm still fairly adamant that those claims are mostly a myth. Because the worst inefficiencies in any workplace has little to do with the individual productivity of the proletariat. And because much of the most valuable work lies in creativity, ingenuity, and subtleties that lies outside of an LLM's expertise.
But loosely speaking, a lot of hacking does in fact come down to a lot of boring, repetitive tasks. Something a machine has always been king of.
The U.S. labs may try to prevent their LLMs to be used for large-scale replicating of all software, hacking, etc, but I don’t understand the view that those things are history - I think they will become incredibly commonplace, especially if there aren’t any cross-industry standards. Do you really believe that it won’t become more commonplace over time?
This will work for approximately a year until ablated open weights models catch up, and then it will be open season. My 2c is that we need to figure out reputation systems that are resilient to low effort cloning rather than try to prevent it, because the latter approach is a non-starter outside of a very short time horizon.
To what extent are ports of old/abandoned games to modern systems permitted? For me, one of the major value propositions of AI-based coding tools is the ability to rapidly update legacy software that still has some use, including games.
Under the expansively vague copyright provisions added thanks to Hollywood lobbyist pressure in the 80s & 90s to deter physical media piracy (Sony Betamax / DeCSS DVD), basically nothing except fair use is clearly "permitted."
However, in practice, a lot of historical retro preservation and emulation for personal use is 'tolerated', as long as it's titles long out-of-print, strictly non-commercial (no one makes money) and kept relatively low-key. Projects like MAME and RetroArch have existed for decades doing truly wonderful work.
While manual decomp / recomps existed in the scene, the 2020 release of the amazing Super Mario 64 decomp SM64 showed that hundreds of hours of painstaking work could unlock ports, deep mods, community translations/expansions, and academic study of classic titles. In the last couple years decomp/recomp has been turbo-charged with LLM agent assistance. Of course, there's now the usual spectrum between carefully guided, curated efforts being accelerated and generally unhelpful slop.
I've been into retro preservation for decades now and the recent explosion of interest, new projects and contributors has been great but there's also concern too much attention, too quickly could attract profiteers who 'cross the line' the community has been careful to avoid, triggering a legal crackdown. So, welcome to the community but please try to not screw up a good thing.
Generally speaking, free emulators without game code or content are fine but free decomps even without original binaries or game assets are a little bit fuzzier. They could be closer to the line but 'how much closer' depends on how it's implemented and the degree to which a company sees it as commercially threatening. So far, there's been no sign of issues but the retro and preservation communities have been careful not to release decomps or recomp ports of titles currently (or recently) for sale. Of course, if someone does a recomp port just for their family and friends, no one will know or care, and if something anonymously ends upon a torrent site, that's just 'software piracy' and on the other side of the firewall. As always, don't talk about or link to that stuff in retro and preservation forums.
I honestly don't see the appeal in cloning existing games. Why not leverage the tools to design and build new games? There are so many mechanics, niches and genres yet to be explored. Use this newly found power and bring them to life.
Imagine there is a game you used to enjoy some time ago, but its servers got shut down, updates changed it in a way you don't like or [insert another N reasons why its not playable anymore]. You don't want a new game, you want to play that exact game.
Now you can just obtain an old build and resurrect it for your own good.
I've worked on private servers for a few games that were shut down, starting long before LLMs were invented. The answer in these cases is that human culture has value to most people. Not only are the gameplay experiences themselves valuable to preserve for new and old players alike, but they also embody periods of time that were in some ways better. As a child, a large amount of my social experiences played out through MMO-type games, which future generations probably won't have, given how worldwide social media controls are effectively requiring people to have an online speech license. Even if such an experience could never legally be rolled out on a large scale again, it's still probably worth holding onto what you can, for smaller communities.
Decomps take the preservation of human culture a step further. In my projects, it has become clear that Windows is no longer a reliable platform, and x86 has a sketchy future, but decompilations can help keep a game alive by moving it somewhere else for the sake of accessibility. They also enable further understanding (whether by AI or people) of the game, along with fixing it and expanding upon it.
Vibing a new experience is valuable, but it's not a substitute for what's already been made.
As a game designer the answer is very obvious: That's hard work!
As a heavy LLM user there is another obvious answer: The LLM is very good at copying.
I suppose those are two sides of the same coin. All the hard decisions and all the creativity is already done.
Those aren't the only options though. I've long told people that want to get into game dev that the best thing to do is mod their favorite games. I'm hopeful that the technical ease to do that now leads to a lot of human creativity being expressed.
I have not explored this but If I wanted Claude Code to make a spiritual successor to Punch Out I would first have it clone Punch Out to see if it was capable of doing this, then if it worked I'd start changing the assets and mechanics one idea at a time until I had something with more novelty behind it.
besides what the others replied, cloning can be done by people with very little skill in the area and in general it's always way easier to work from a proven design (existing game) than to come up with one from scratch.
This is why we need open AI (not OpenAI). Some duopoly deciding what goes on a whim is not acceptable
I can sorta see them suppressing instructions for say bioweapons (that seems fair to me), but this demonstrates that they'll bend for corporate interests.
> preemptively protecting themselves from potential lawsuits ... for offering the means to reverse-engineer their software.
I think the legal concern isn't so much the act of reverse engineering itself which is mostly legal (except RE specifically to remove anti-copying protection under DMCA), it's that damages from contributory infringement are theoretically unbounded because it's up to $150k per work and $2,500 per download. When the client is worth billions, corporate counsel is going to default maximally conservative, which is extra tragic because secondary infringement almost certainly doesn't apply to AI in this context. Meanwhile, overreacting to this extremely unlikely threat IS causing significant disruption to innocent activities.
Under copyright law secondary infringement doesn't apply to "general purpose technology". It requires "encouraging or knowing in advance it'll be used for infringement," set in the early 80s in the Sony BetaMax case. Under DMCA it requires being "primary designed to circumvent access controls," set in the late 90s by Hollywood lobbyists targeting DeCSS for DVDs. Lawyers today are only defaulting to prohibition because there hasn't yet been a case ruling AI is a general purpose technology. Yet it's obviously the most 'general purpose' tech since, maybe.... the wheel?
OAI / ANT management needs to revisit the secondary copyright infringement prohibition because it's hurting not only their product and brand but also the entire AI value prop. A helpful intelligent agent that randomly goes AWOL for irrationally mistaken reasons is a serious product issue. While mistaken concerns around bioweapons and IT security are understandable, dropping secondary copyright infringement prohibitions seems like an easy win.
The real liability to OAI/Ant is much smaller than it appears, while the long-term downsides are larger than they look. Pragmatically, copyright trolls are just a nuisance, the real threat is someone with deep pockets like Nintendo/Sony, but those guys don't want to sue anyone who might go to court. The last thing they want is a clear ruling that AI is general purpose which would forever remove their vague threat.
They never expected their models could be used for reverse engineering.That's the truth.
Now, they're panicking. And this is why the other dude, OpenAI, came out to say that people shouldn't use RAM locally, because they can't control the software anymore.
Open weight models, ablated models are going to become the norm, hopefully, locally run.
So instead, they're saying, look, hardware should become hard to reach for all of us, for the consumers.It's not because we're competing for resources for the same silicon or chips as they are.The ship has sailed thankfully.
One of the new threat models of local LLMs, especially abliterated ones, is hidden instructions or activation words that obscured trigger tool calls which exfil data.
Which is why agents should always run in a sandbox with restricted network access
An agent running in a restricted sandbox helps fuck-all if the underlying prompts are sent to a mothership datacentre somewhere.
Did you miss the word ‘local’ above?
And how do you suppose that happens when it's running in a sandbox that prevents things like network access?
For people touting about how their software can destroy the world, they aren't awfully creative on how it can achieve that. Hacking breaches of IP was probably most cynicists' top 3 questions asked when presented with the idea of "boosting productivity". If you make it "10x" faster to write code, you make it 10x faster for a hacker to hack. Naive cryptographic principles are less about an ironclad security and more about encryption that takes too much time to brute force. It all leads to one thing.
I'm still fairly adamant that those claims are mostly a myth. Because the worst inefficiencies in any workplace has little to do with the individual productivity of the proletariat. And because much of the most valuable work lies in creativity, ingenuity, and subtleties that lies outside of an LLM's expertise.
But loosely speaking, a lot of hacking does in fact come down to a lot of boring, repetitive tasks. Something a machine has always been king of.
remember when llms were a tool and you could use them however you wanted? pepperidge farms remembers
can't wait for the day when you end up in jail for the wrong prompt or when you will be served ads through the kazoo.
The U.S. labs may try to prevent their LLMs to be used for large-scale replicating of all software, hacking, etc, but I don’t understand the view that those things are history - I think they will become incredibly commonplace, especially if there aren’t any cross-industry standards. Do you really believe that it won’t become more commonplace over time?
> can't wait for the day when you end up in jail for the wrong prompt
I also can't wait for yesterday.
[1] https://www.smh.com.au/national/nsw/sydney-man-allegedly-dis...
[2] https://www.theverge.com/ai-artificial-intelligence/1004747/...
[3] https://www.washingtonpost.com/technology/2026/08/27/chatgpt...
That’s not really an LLM thing though - if she made the same comment on a Facebook post, she’d face the same consequences.
Is there no difference between posting something publicly and prompting a tool privately? Intent seems different to me.
I’ve said some pretty obscene things to Siri out of frustration and haven’t gotten arrested yet.
This will work for approximately a year until ablated open weights models catch up, and then it will be open season. My 2c is that we need to figure out reputation systems that are resilient to low effort cloning rather than try to prevent it, because the latter approach is a non-starter outside of a very short time horizon.
Theyre already here. Theyre slower, but if you have 128GB And watts+time, you can so it.
Imvestors should be better. We dont need new models to do this, its now the hardware cartel propping up prices. Break that dam and moats gone Sam.
Nvidia, openai, anthropic; they know it now. 1 "frontier" model is for the 1%. Is equal to 10 models for ths majority.
https://news.ycombinator.com/item?id=50304358
[dead]
To what extent are ports of old/abandoned games to modern systems permitted? For me, one of the major value propositions of AI-based coding tools is the ability to rapidly update legacy software that still has some use, including games.
> ... permitted?
Under the expansively vague copyright provisions added thanks to Hollywood lobbyist pressure in the 80s & 90s to deter physical media piracy (Sony Betamax / DeCSS DVD), basically nothing except fair use is clearly "permitted."
However, in practice, a lot of historical retro preservation and emulation for personal use is 'tolerated', as long as it's titles long out-of-print, strictly non-commercial (no one makes money) and kept relatively low-key. Projects like MAME and RetroArch have existed for decades doing truly wonderful work.
While manual decomp / recomps existed in the scene, the 2020 release of the amazing Super Mario 64 decomp SM64 showed that hundreds of hours of painstaking work could unlock ports, deep mods, community translations/expansions, and academic study of classic titles. In the last couple years decomp/recomp has been turbo-charged with LLM agent assistance. Of course, there's now the usual spectrum between carefully guided, curated efforts being accelerated and generally unhelpful slop.
I've been into retro preservation for decades now and the recent explosion of interest, new projects and contributors has been great but there's also concern too much attention, too quickly could attract profiteers who 'cross the line' the community has been careful to avoid, triggering a legal crackdown. So, welcome to the community but please try to not screw up a good thing.
Generally speaking, free emulators without game code or content are fine but free decomps even without original binaries or game assets are a little bit fuzzier. They could be closer to the line but 'how much closer' depends on how it's implemented and the degree to which a company sees it as commercially threatening. So far, there's been no sign of issues but the retro and preservation communities have been careful not to release decomps or recomp ports of titles currently (or recently) for sale. Of course, if someone does a recomp port just for their family and friends, no one will know or care, and if something anonymously ends upon a torrent site, that's just 'software piracy' and on the other side of the firewall. As always, don't talk about or link to that stuff in retro and preservation forums.
That's what happens when you rent your intelligence from someone else.
[dupe] Discussion: https://news.ycombinator.com/item?id=50304358
I honestly don't see the appeal in cloning existing games. Why not leverage the tools to design and build new games? There are so many mechanics, niches and genres yet to be explored. Use this newly found power and bring them to life.
Imagine there is a game you used to enjoy some time ago, but its servers got shut down, updates changed it in a way you don't like or [insert another N reasons why its not playable anymore]. You don't want a new game, you want to play that exact game.
Now you can just obtain an old build and resurrect it for your own good.
I've worked on private servers for a few games that were shut down, starting long before LLMs were invented. The answer in these cases is that human culture has value to most people. Not only are the gameplay experiences themselves valuable to preserve for new and old players alike, but they also embody periods of time that were in some ways better. As a child, a large amount of my social experiences played out through MMO-type games, which future generations probably won't have, given how worldwide social media controls are effectively requiring people to have an online speech license. Even if such an experience could never legally be rolled out on a large scale again, it's still probably worth holding onto what you can, for smaller communities.
Decomps take the preservation of human culture a step further. In my projects, it has become clear that Windows is no longer a reliable platform, and x86 has a sketchy future, but decompilations can help keep a game alive by moving it somewhere else for the sake of accessibility. They also enable further understanding (whether by AI or people) of the game, along with fixing it and expanding upon it.
Vibing a new experience is valuable, but it's not a substitute for what's already been made.
As a game designer the answer is very obvious: That's hard work!
As a heavy LLM user there is another obvious answer: The LLM is very good at copying.
I suppose those are two sides of the same coin. All the hard decisions and all the creativity is already done.
Those aren't the only options though. I've long told people that want to get into game dev that the best thing to do is mod their favorite games. I'm hopeful that the technical ease to do that now leads to a lot of human creativity being expressed.
I have not explored this but If I wanted Claude Code to make a spiritual successor to Punch Out I would first have it clone Punch Out to see if it was capable of doing this, then if it worked I'd start changing the assets and mechanics one idea at a time until I had something with more novelty behind it.
Cloning existing games can be done with little to no work.
Building something new is challenging, with or without AI its orders of magnitude slower than copying.
Building something new is also risky. Copying something with proven vitality is less so.
Given this almost tautologically you should expect way more clones than novel things.
besides what the others replied, cloning can be done by people with very little skill in the area and in general it's always way easier to work from a proven design (existing game) than to come up with one from scratch.
Bait
Nostalgia
- all those RE guys ll just hop to chat gpt instead
- if chat gpt also blocks it, they ll hop to all the chinese models
You say that like it’s a problem? Sounds like the junkies have to be fed one way or another.
But copying Ghibli style is still allowed.
Software people protecting their "own" and fucking over everyone else.