Epic's Tim Sweeney says AI labels on Steam are meaningless as generative tools become universal

Skye Jacobs

Posts: 1,913   +58
Staff
A hot potato: Epic Games CEO Tim Sweeney is pushing back against the way major PC storefronts flag games that use generative AI. He argues that labels like Steam's "AI-generated content disclosure" have no place in online game stores because AI will soon be embedded in almost every part of the production process.

Valve introduced a formal policy on generative AI for Steam in early 2024, requiring developers to disclose how AI systems are used in each game and adding visible 'AI-generated content' disclosures on individual store pages. Nexon's co-op shooter Arc Raiders is one such example.

Sweeney's position, laid out in a reply on X, is that this kind of tagging might make sense for galleries or stock-asset marketplaces where authorship and licensing status are central to the transaction, but not for consumer game stores. In his view, generative AI is quickly becoming just one more tool in a long chain of software – from compilers to physics engines – that underpins game development. Once AI is ubiquitous, singling it out on store pages becomes meaningless.

The Epic CEO has previously argued that generative systems can shrink production bottlenecks and give smaller teams access to capabilities that once required large art, writing, or QA departments. He has described scenarios where models generate context-sensitive dialogue guided by human voice actors rather than entirely replacing them.

The disagreement over labels comes amid unresolved legal fights about whether training data scraped from the public internet infringes the rights of artists, writers, and other rights holders whose work ends up in model corpora without permission or compensation.

Media companies have begun suing AI vendors over alleged unauthorized use of copyrighted catalogs, and rights experts expect those cases to shape how courts treat both model training and downstream outputs. Valve's disclosure rules are one way platforms are trying to insulate themselves from that uncertainty.

For now, PC storefronts are experimenting with different answers to the same question: whether generative AI is a special case that warrants consumer-facing labels or simply another layer in the tech stack that can be handled behind the scenes.

If Sweeney's prediction proves accurate and AI appears in nearly every release, platform-level disclosures could eventually become so common as to be little more than a checkbox. However, many developers and players still argue that the label matters because it encodes both ethical concerns about training data and aesthetic preferences about how games should look and sound.

How long the gap between Steam's approach and Epic's will persist may depend less on technical adoption – AI use in game development is expanding rapidly. Instead, it may hinge on whether regulation, litigation, or sustained consumer pressure forces game stores to treat AI not just as an internal production detail, but as material information that belongs on the front of the box.

Permalink to story:

 
I disagree. Transparency is the way to go forward when it comes to AI. Even I am using it for programming, but that's considered normal now. You just HAVE to understand the code or else you are just blindly trusting the AI.

It's such a pain the ars to correct an AI when you know it made a mistake: "I'm sorry, you are right, I'll correct it." - proceeds to write the same thing or something worse *nerd*
 
Last edited:
Ok, so when AI is in everything Valve can change the disclosure requirement again?

Until then why not have the disclosure in place and let people decide where their money goes?
 
It's such a pain the ars to correct an AI when you know it made a mistake: "I'm sorry, you are right, I'll correct it." - proceeds to write the same thing or something worse *nerd*
So annoying when it does that, you correct it, show it the right way to do it and it will shower you with praise saying you're right and then 2 prompts later when you want something altered it's still ittirating on its own code rather than what you showed it that works.
 
So annoying when it does that, you correct it, show it the right way to do it and it will shower you with praise saying you're right and then 2 prompts later when you want something altered it's still ittirating on its own code rather than what you showed it that works.

I think a local AI model is better then using any of those paid ones in my opinion. At least you get to tweak, tune and train it the way you want.

I often get borked results either breaking code or suddenly inserting chinese stuff into my code - as if there was a collision somewhere with my instance vs someone else.

I can't wrap my head around it either the billions being poured into, it still is not capable of writing a common, normal text or bunch of paragraphs without making mistakes or "signs" that it was obviously AI generated.

Have you noticed the word "Tailored" all over the place lately? Well there you have it. As if it was a trademark for search engines to understand it was generated by Ai.
 
Another day, another whining by Sweeney.

Why can't this dude just sit on his mountains of money with his mouth shut?
 
Tim Sweeney, like so many "enlightened thinkers" of our time, is conflating "availability" with "utility": that because AI―although readily available in the present and being far too commonly used, where it has no business existing, primarily because of its predilection to fabricate information whole cloth way too often―is soon going to be as mundane a fixture in our lives as a Starbucks corner store, that that is somehow beneficial to the end user. The thinking really does seem to be "if there's nowhere it can't be found, then complaining about its presence is about as useful as yelling at clouds. Just submit already! 'The future is now', old man!" To which I say, "That's just, like, your opinion, man."

ChatGPT might be able to summarize why the shape of a baboon's *** makes for a good cup holder, but "no" is a complete sentence. Consent matters.
 
It seems like some people want to support games that don’t use AI, so Valve could gain sales from those people and Epic won’t. If you want the sale, you should give the people what they want. Steam is the only one who gets it and that is why people love them. Take care of your customers and they will take care of you. Most game stores treat their customers like they are just a bunch of pirates who need to be domesticated while harvesting and selling their data in the background as they complain about the bad monopolists they allege are Steam.
 
Back