Saturday, October 31st 2026

Anthropic Bans AI Game Clones and Reverse Engineering With Claude

AI game cloning, reverse-engineering, and decompiling have all become increasingly popular of late, with a cottage industry of gaming projects popping up almost overnight, including everything from controversial Fortnite and Fallout clones to fan-made console ports. Now, it seems as though Anthropic may have had enough of the bad press and potential pressure from gaming industry giants, with some users reporting (via Tom's Hardware) that Claude will no longer recreate game mechanics or existing games.

One X user flagged a new rule in Claude that could potentially be the source of the restriction, which states "Do build original games, games in a genre, and games from the user's own designs and assets. Don't build a recreation of a specific existing game, even if it is renamed or called original." While the intention with the new rule may have been to prevent copyright infringement and potential legal implications for Anthropic, it seems as though it has resulted in a flurry of new issues on the Claude GitHub, with reports including claims that Claude is refusing to clone a developer's own work from earlier in a project for a similar mechanic in a latter part of the game.
Other users on sites like Reddit have reported that the issue extends beyond just gaming, with reports citing a refusal to use the popular Reverse Engineer Anything, which has been used to do everything from break free of manufacturer locks on smart TVs to adding features to car infotainment systems, and even recompile video games for different platforms.
Sources: @chasmmmmmmmmmmm on X, via Tom's Hardware
Add your own comment

18 Comments on Anthropic Bans AI Game Clones and Reverse Engineering With Claude

#1
mb194dc
You'd have to be crazy to reverse engineer protected ip on a cloud model to start with. Another win for local.
Posted on Reply
#2
Shihab
Hol'up! You're telling me that the revolutionary technology that will make programmers obsolete and every lay person can now make any software they want, actually can't be used to make any software the lay person wants?

No really, has there been any single good headline about those idiots lately? Junk credit ratings, malware dissimination, andd now self-inflicted product degredation? Just die already!
Posted on Reply
#3
sounreal
Saw this coming 100 miles away.
Posted on Reply
#4
Darmok N Jalad
It was a matter of time. There’s already an alpha suite of AI-reverse-engineered Adobe clones out there. I’m waiting for them to get the take down notice. It’s okay for models to steal from the entirety of the internet, but it dare not go after the big guys who have legal teams.
Posted on Reply
#5
Steevo
Darmok N JaladIt was a matter of time. There’s already an alpha suite of AI-reverse-engineered Adobe clones out there. I’m waiting for them to get the take down notice. It’s okay for models to steal from the entirety of the internet, but it dare not go after the big guys who have legal teams.
Some animals are more equal than others.
Posted on Reply
#6
Aretak
Mass copyright infringement for me, but not for thee.
Posted on Reply
#7
Tess
sounrealSaw this coming 100 miles away.
Yep, I even said so here on TPU a few weeks ago.

Local modelswill take it upon themselves to continue it though.
Posted on Reply
#8
GvonTech
It should be said that they have made an attempt at banning the practice but as we all know there are still ways to circumvent such guardrails so they really have done nothing but given themselves the excuse of, "We did what we could."
Posted on Reply
#9
Let3r
SteevoSome animals are more equal than others.
Some AI models are more equal than others.
Posted on Reply
#10
Grimling
These cloud-based AI solutions will be nothing more than Government-Propaganda-Tools (GPT) in the future.
If you want to actually have a useful AI, that is doing what you telling it to do, then you will need to have it locally running and open source!
Posted on Reply
#11
lexluthermiester
This move will not stop reverse engineering, it will only slow it down and shift it to offline AI.
Posted on Reply
#12
jdrch
mb194dcYou'd have to be crazy to reverse engineer protected ip on a cloud model to start with. Another win for local.
You need probably 1 TB of RAM (a fortune even pre-AI) to get to 50% of the performance of a frontier cloud model. ChatGPT will probably still do the trick, though
Posted on Reply
#13
Darmok N Jalad
jdrchYou need probably 1 TB of RAM (a fortune even pre-AI) to get to 50% of the performance of a frontier cloud model. ChatGPT will probably still do the trick, though
The question becomes a matter of economics. To use a model that needs that kind of memory footprint, what should it cost the consumer where the host actually turns a profit? The spend is drastically outpacing revenue, currently. Curious if/when what has been spent will ever get to break-even.
Posted on Reply
#14
mashie
jdrchYou need probably 1 TB of RAM (a fortune even pre-AI) to get to 50% of the performance of a frontier cloud model. ChatGPT will probably still do the trick, though
Not really, Deepseek v4.1 Flash 1017 needing 512GB VRAM/unified RAM and pretty close to GPT-5.6 in capability.
Posted on Reply
#15
evernessince
lexluthermiesterThis move will not stop reverse engineering, it will only slow it down and shift it to offline AI.
Might not even slow it down. Chinese models are right behind them and offer uncensored models at a fraction of the cost.
Posted on Reply
#17
lexluthermiester
evernessinceMight not even slow it down. Chinese models are right behind them and offer uncensored models at a fraction of the cost.
Good point!
Posted on Reply
#18
jdrch
Darmok N JaladThe question becomes a matter of economics. To use a model that needs that kind of memory footprint, what should it cost the consumer where the host actually turns a profit? The spend is drastically outpacing revenue, currently. Curious if/when what has been spent will ever get to break-even.
From what I understand, this is why the top tiers of Claude, etc. are $200/month. The majority of current spend is on overbuying hardware that's sitting in warehouses, not on query compute per se.

That said, AFAIK none of the major AI players have public facing data about how profitable each tier is. The only data I've seen is from Google's latest quarterly report in which they said AI search queries make less revenue than their conventional search did, but that doesn't say anything about cost or margin.
Posted on Reply
Add your own comment
Nov 20th, 2026 13:11 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts