Skip to content
trending

Magazine

Call of Duty: Black Ops 7 and the AI Strike Nobody Noticed

When Call of Duty admitted to using AI-generated assets, it revealed how artificial intelligence entered gaming through the back door - formally disclosed but practically invisible.

By 8 min read
Call of Duty: Black Ops 7 and the AI Strike Nobody Noticed

👁 113 reads

On November 14, 2025, the day Call of Duty: Black Ops 7 launched, players started taking screenshots. Not of gameplay - of calling cards, the small images you use to decorate your profile. Fingers that didn’t quite add up. Faces slightly misaligned with the pattern. One graphic in the style of Studio Ghibli, looking exactly like something from a generator.

The answer required no investigation. You just had to scroll down the game’s Steam page to a section that didn’t exist three years ago: “Our team uses generative AI tools to help create some assets in the game.” The world’s biggest entertainment franchise admitted it in a field it filled out itself, and nobody noticed until launch.

Activision responded to the wave of criticism with a statement that already sounds like a ready-made formula in this industry.

Like many people around the world, we use various digital tools, including AI tools, to enhance and support our teams in creating the best experiences for players. Our creative process is still led by talented people in our studios.

Activision statement after Black Ops 7 launch

The company didn’t specify which elements were machine-generated. It didn’t have to - the Steam declaration only says “some assets,” and nobody has a tool to verify it.

This is the entire story of artificial intelligence in games in one paragraph: the technology entered through the side door, formally disclosed, practically invisible, until someone’s fingers didn’t add up.

A Year Without Voice

Before players started counting fingers on calling cards, the industry went through the longest labor dispute in its history. The American actors’ union SAG-AFTRA began a strike against game producers on July 26, 2024. It wasn’t mainly about rates - it was about digital copies. About whether an actor’s recorded voice and motion could be used to generate lines that actor never spoke.

The strike lasted nearly a year. A tentative agreement was reached on June 9, 2025, and members ratified it on July 9 with a vote of 95.04 to 4.96 percent - that’s not a negotiation result, that’s a referendum on something obvious.

The agreement gave actors three things. The obligation to obtain consent to create a digital replica. The obligation to inform that such a replica will be used. And - cleverly - the right to suspend once-given consent to generate new material during a strike, so producers couldn’t wait out the protest by playing scenes with strikers’ voices. Plus raises: 15.17 percent from ratification and three percent each in November 2025, 2026, and 2027.

The actors won. But they won a contract about their voices - not a principle that games should be made by humans.

One in Five Games

The scale of the phenomenon is best seen on Steam, because Valve has required disclosure since January 2024. By mid-2025, about eight thousand titles had filed it. A year later it’s a different reality: a review from July 28, 2026 showed that roughly one in five games in the store’s content survey carries such a label. In individual weeks it gets more dramatic - the portal PCGuia calculated that in the week ending June 19, 2026, forty percent of Steam releases had declared generative AI.

Valve rewrote the form along the way. Since January 2026, disclosure isn’t required for tools that only assist production itself - code-writing assistants fell out of the requirement. Two categories remain that players actually see: content generated in advance and inserted into the game, store page, or marketing materials, and content generated live during gameplay.

This distinction is more important than it looks. The first category is a matter of honesty with buyers. The second is a preview of games where character dialogue won’t exist until you trigger it.

Who’s Already Signed On

Major publishers stopped hiding it around fall 2025. Electronic Arts announced a strategic partnership with Stability AI - the company behind Stable Diffusion - to jointly build models and tools for their own artists. The first concrete result is supposed to be accelerating the creation of PBR materials, textures that describe how surfaces reflect light. It’s not flashy, but this is exactly the kind of thing entire departments sit on.

Microsoft showed the Muse model, which generates gameplay itself based on what the player does. Ubisoft at its Paris studio, together with Nvidia and Inworld, is developing the NEO NPC prototype - an independent character you can talk to freely, instead of choosing lines from a list. A prototype, not a product; but the direction is clear.

Korean publisher Krafton, which publishes PUBG, put it most openly. The company announced a “complete reorganization” toward becoming an “AI-first” enterprise and pledged over 130 billion won, about 88 million dollars, for agentic AI infrastructure from 2026. It halted recruitment outside areas related to artificial intelligence.

A few weeks after this declaration, Krafton launched a voluntary departure program for employees in Korea, assuring it wasn’t a workforce reduction. Players received it unambiguously and calls for boycott appeared. Three months later, the newly appointed AI director explained that the company “uses AI as a tool that enhances human imagination and creativity, not replaces it.”

This sequence - declaration, layoffs, reassuring clarification - repeats regularly enough that it’s becoming a literary genre. Microsoft, Sony, Take-Two, and Embracer reduced employment, and AI appeared in justifications as a factor shortening production cycles. At the same time, reports from major publishers say implemented tools frustrate developers more often than they actually relieve them. Both things can be true at once - and usually are.

Parallel to the corporate dispute runs a second, uglier one, in a place no labor union covers: in mods.

Modding communities - most vocally around Skyrim - started cloning actors’ voices with tools like ElevenLabs to add lines to characters that weren’t in the game. Many of these modifications are pornographic content. Actors spoke unambiguously that they didn’t consent to this and consider the practice abuse. The National Association of Voice Actors put it briefly:

The harm done to voice actors and gaming companies by AI and synthetic voices is real and measurable. Voice actors have no legal tools to fight this.

National Association of Voice Actors

The largest mod platform, Nexus Mods, stated that AI-generated content doesn’t violate its terms of service, though it provides a form for reporting rights violations. The effectiveness of this path is, to put it mildly, debatable.

The contrast with the other pole is striking. Michael Caine and Matthew McConaughey signed licensing agreements with a speech synthesis company for their own voices - consciously, for money, on their terms. The same technology, two completely different worlds. The difference is called consent.

A Voice That Returned

The most interesting thread of this story is older than the whole controversy and doesn’t fit simple divisions at all.

Miłogost Reczek, one of Poland’s best voice actors, died in 2021. In Cyberpunk 2077 he was the voice of ripper doc Viktor Vektor - a character you can’t get through the main storyline without. When CD Projekt Red was preparing the Phantom Liberty expansion, it faced a choice: hire another actor and re-record all of Viktor’s lines from scratch, or drop the character.

The studio chose a third path. It recorded a replacement actor, then ran those recordings through Respeecher software to sound like Reczek. The family gave consent - as Bloomberg reported, the actor’s sons were very helpful in this process. Viktor Vektor’s voice in the 2023 expansion is the voice of a man who was no longer there.

You can call it technological necrophilia or tribute, depending on which side of the dispute you stand on. I lean toward the latter, because you can see everything that’s missing in most cases: conscious consent from those entitled, transparency, and a purpose that wasn’t cost-cutting.

“We Should Have Just Let You Know”

Two years later, the industry got a case from the other pole, and from a studio with the best reputation.

11 bit studios’ The Alters released on June 13, 2025 and received very good reviews. A week later, players started finding things in the game they didn’t expect: a fragment of background text on a graphic, looking like raw output from a generator complete with a leftover prompt, and subtitles for videos translated by machine.

The studio admitted it quickly. In a statement, it wrote that the text on the graphic was a temporary placeholder inserted by an artist and was supposed to disappear before launch, but didn’t due to internal oversight. Video translations were commissioned to AI because time ran out before launch, with the intention of passing them to trusted localization agencies later - which they did after the controversy.

We should have just let you know.

11 bit studios in statement about use of generative AI in The Alters

This sentence is basically the entire ethics of this industry in a nutshell. Nobody fired the localization team. Nobody sold a generated game as handmade. One thing was missing: saying it outright before someone found it.

Forty Prototypes in a Week

What does this look like from the perspective of someone running one of Europe’s largest studios? Michał Nowakowski, one of CD Projekt Red’s executives, told Edge magazine about a conversation with the founder of a studio based almost entirely on AI.

In a week I can have 40 prototypes, in two weeks I can have five games that I consider the best, and after three weeks I actually release a game.

AI studio founder’s statement, quoted by Michał Nowakowski

Nowakowski doesn’t question that such games will be made - he’s certain of it. However, he doubts whether it’s the right path, and bets on what he calls a fresh idea with soul. CD Projekt itself runs several dozen research projects related to AI, including developing its own models. Techland uses artificial intelligence for NPC behavior and dynamic dialogues.

In other words: studios aren’t standing aside. They’re doing exactly what the rest of the world does - using AI in the middle of the production process, where players don’t see it, and speaking very carefully about it publicly.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Who Will Notice

Artificial intelligence won’t take away from the gaming industry what’s most expensive in it: the idea, direction, that hard-to-name sense that makes one game engaging and another not. However, it will take work from people in departments whose names nobody remembers - artists for secondary materials, translators, supporting actors, creators of icons and calling cards.

And here’s the core that’s invisible in ethics disputes. Players detected AI in Black Ops 7 not because they have sharpened consciences - but because the graphics were simply worse. As long as generated things are poor quality, the public will catch them and the controversy will work as a quality control mechanism. When they stop being poor quality, nobody will protest, because there’ll be nothing to notice.

That’s why from this whole story I hold onto the case of Viktor Vektor most. Not because it was technology better than others - exactly the same, the same Respeecher, the same mechanism. Because someone called the family and asked.

Kinetyra

The link to ElevenLabs is a partner link: if you buy through it, QUANTSLANT earns a commission and you pay the same price. It has no influence on the content, including where the coverage is critical.

// read next

More on this from QUANTSLANT

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

read nextI Tested Claude Fable 5.1 - One Task Froze ItbenchmarksI Tested Claude Fable 5.1 - One Task Froze ItWe ran Claude Fable 5.1 through language tests on launch day. The new model hung on a simple grammar task and never returned an answer.

Most read

  1. 1Best AI Video Generator Apps for Your Phone (2026)
  2. 2Claude Opus 5.5 Motion Graphics: 60s Showreel from One Prompt
  3. 3Wan 2.2 on RTX 4090: timings, VRAM and four test clips
// site map

This page in other languages6

Quantslant in numbers 1,427posts in archive25news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›