Three studios got caught using AI this week — and all gave the same answer

Gaming 4 min 80 sources

Sega, Crystal Dynamics, and the Assassin's Creed co-creator's studio all disclosed generative-AI assets within days of each other. The near-identical scripts they used to explain it reveal more than the tech does.

Key takeaways

Three game studios — Sega, Crystal Dynamics, and the Assassin's Creed co-creator's studio — disclosed using generative AI within one week, each calling it "just a tool" while insisting "real humans" made the final work.
When competing studios all reach for the same defense under the same pressure, the uniformity reveals a shared incentive — cut cost and time — not an honest account of how much AI was used.
The disclosures exist because Steam now requires them, and their vagueness lets a single AI asset and most-of-the-game read the same on the page.

In the space of one week, three separate game studios disclosed they had used generative AI to make parts of their games. What’s striking isn’t the AI. It’s that all three reached for almost the same words to explain it.

What happened

At the Xbox summer showcase, Sega revealed a new Crazy Taxi: World Tour — the first mainline entry in over two decades. Within hours, players found a disclosure on its Steam page: the game uses generative AI “as a support tool for developers” [1][10]. The franchise’s original creator, Kenji Kanno, defended it, saying artists used AI to “generate ideas” and then “draw the actual thing” — “everything is made by an actual human” [1].

Days earlier, Crystal Dynamics confirmed its Tomb Raider: Legacy of Atlantis remake used gen AI — calling it “a tool that can help our teams” get “the right answers faster,” with all final assets human-made [7][22]. And Panache Digital Games — the studio founded by Patrice Désilets, co-creator of Assassin’s Creed — apologized after players spotted AI-generated art in the free prologue of 1666: Amsterdam, including in-game portraits and marketing key art. The studio said human-made versions would replace them “soon” [3][5].

The system underneath: a convergent script

Read the three statements side by side and a template appears. AI was “a support tool.” It made “early versions” or “references,” not the final thing. “Actual humans” did the real work. The assets were a placeholder oversight, now being fixed.

These studios compete with each other. They are not coordinating. Yet under the same pressure — a player backlash that threatens wishlists, the list of people who flag a game to buy at launch — they converge on the same handful of phrases. When rivals who’d normally differentiate all say the identical thing, you’re usually not hearing an explanation. You’re hearing the shape of the incentive they all share.

The incentive is concrete. AI assets cut the cost and time of producing concept art, textures, portraits, and signage. A studio that ships faster and cheaper protects its margin. But players increasingly treat AI art as a sign of cut corners — one 1666: Amsterdam Steam reviewer wrote, “I’m de-wishlisting it and ignoring the company” [3]. So the disclosure becomes a defense: admit the cheaper method, but frame it as marginal and reversible, and promise the “real” version is coming.

What changes — or is now in motion

A pattern is hardening. As Kotaku noted, most prior cases were studios using AI in concepting and then “frantically apologizing when some AI art made it into the final product,” or trying to “sneak it in without people noticing” [10]. Sega’s case is different — it’s one of the first times a big studio has disclosed AI in shipped in-game assets up front, without apology [10].

That up-front disclosure is itself a product of pressure: Steam now requires developers to declare AI use on the store page. The disclosure exists because a platform forced it, not because studios volunteered transparency. And the vagueness is doing work — Sega’s wording is loose enough that, as Kotaku put it, “one could assume anything from a single asset being made with AI to most of the game” [10].

The counter-voices

Not everyone is reaching for the script. A Plague Tale studio Asobo pushed back directly: smaller teams don’t need gen AI to compete — “it’s a choice,” not a necessity [2]. And an ex-Valve veteran, Left 4 Dead and Half-Life writer Chet Faliszek, argued in a widely-shared clip that a game like Team Fortress 2 — built on specific, intentional design choices — could never have come from AI [12]. The disagreement isn’t really about whether AI works. It’s about whether “faster and cheaper” is the same thing as “better,” and who decides.

The angle, for a player

This is why a game can look slightly off in ways you can’t name — a portrait with too-smooth hands, signage that reads as nonsense up close. The model that rewards shipping faster and cheaper is now pointed at art the way live-service pointed studios at endless patches. The disclosure on a Steam page isn’t transparency winning. It’s a platform rule meeting a PR script, and the gap between “a tool” and “most of the game” is exactly the part nobody is required to fill in.

02 · Lesson · why it matters

When rivals all give the same answer, listen to what they share

The moment competitors who should disagree start using identical words, you're not hearing the truth — you're hearing the pressure they all stand under.

Three studios, one sentence

This week three game studios admitted to using generative AI. They don’t work together. They make different games for different audiences and compete for the same players’ money. Yet when each one explained itself, the words came out nearly the same: AI was “just a tool.” It made “early versions,” not the real thing. “Actual humans” did the work that matters.

When people who have every reason to differentiate suddenly sound alike, that sameness is itself the signal. It tells you something they all have in common is doing the talking.

It’s easy to assume identical messages mean a secret meeting. There wasn’t one. Sega, Crystal Dynamics, and a small studio in Montreal did not call each other to agree on a line.

What they shared was a situation. Each had used AI to make art faster and cheaper. Each then faced the same threat — players who treat AI assets as a sign of a corner cut, and who pull their interest before launch. Under one shared pressure, three separate actors arrived at the same defense. Not because they planned it. Because the pressure was the same, and there was only one safe-looking way out.

That’s worth holding onto. You don’t need a conspiracy to get a chorus. You just need a lot of people standing in the same trap, each reaching for the same exit.

The words are doing a job

Look at what the shared phrases accomplish. “Just a tool” makes the AI sound small. “Early versions” makes it sound temporary. “Real humans did the final work” moves your attention to the part that’s reassuring and away from the part that isn’t.

None of those phrases is a lie, exactly. AI probably was used as a tool. Some assets probably were early. Humans were probably involved somewhere. But the words are chosen to leave a gap — between “a single asset” and “most of the game” — and to let you fill it with the kinder guess. The job of the script isn’t to inform you. It’s to manage how you feel while staying technically true.

Why this matters past games

This pattern is everywhere once you see it. Airlines explaining a fee. Banks explaining a charge. Three companies in an industry all describing a layoff with the same phrase about “aligning resources.” The tell is never one statement. It’s the repetition across people who, by rights, should be telling you different things.

When you hear it, the useful question flips. Not “is this statement true?” — it usually is, narrowly. The question is: what pressure do all these speakers share, that makes this the only thing they can safely say? Answer that, and you understand the situation far better than any of the speakers will tell you directly.

Where you stand in it

Here’s the part that should keep you humble rather than just sharp. You are also someone who reaches for a safe, shared script when you’re under pressure — the polite reason you give for canceling, the standard line everyone uses to explain being late, the phrasing your whole team adopts when a project slips. You converge too. Not because you’re dishonest, but because a shared bind narrows everyone’s words the same way.

So seeing the script in three studios isn’t a reason to feel above them. It’s a reminder that the script is a feature of being squeezed, and almost everyone gets squeezed. The skill isn’t catching liars. It’s noticing when a situation — yours or theirs — has quietly written everyone’s lines, and remembering how little any single line can tell you about how much was really going on behind it.

03 · Lab · your turn

Read the Room

Rehearse spotting when rival voices converge on one script, and discover the shared pressure behind it by switching that pressure off.