Daybreak - Microsoft called Copilot "entertainment only." Then killed it on Xbox

Episode Date: May 26, 2026

In 1998, a Metal Gear Solid villain named Psycho Mantis read your memory card out loud and made your controller vibrate on its own. Players were stunned. It felt like a genuine invasion. And they loved it.In 2026, Microsoft built an Xbox assistant that could do roughly the same thing. Plus some more. Track your history, read your screen, coach you through the game. Players were stunned. It felt like a genuine invasion. And they hated it.The viral hate train began in March 2026. Two months later, the new Xbox CEO killed it.The backlash wasn't really about the technology. It was about what the technology misunderstood. Game design requires a careful balance between challenge and ease that makes it worth playing. And an AI assistant wasn't really reducing friction, it just introduced a different, kind of insulting type.So where does AI actually belong in gaming?Daybreak is produced from the newsroom of The Ken, India’s first subscriber-only business news platform. Subscribe for more exclusive, deeply-reported, and analytical business stories.

It's 1998. You're playing Metal Gear Solid on your first-generation PlayStation sitting very close to a boxy TV in a dark room. Midway through the game, you encounter a villain named Psycho Mantis. He has telekinetic and telepathic powers and he tells you he knows everything about you. To prove it, he breaks the fourth wall. The music shifts and he begins slowly to read the names of the other games you have saved on your memory card. The ones you've been playing over the last week, the last month. He says things like,

I say you like Castlevania. So you're like Suikodin. You enjoy role-playing games. He is levitating off the ground menacingly. And once he's done reading your card, he asks you to put down your controller so he can show you how powerful he really is. Almost immediately,

the controller starts vibrating violently all on its own. And across living rooms around the world, players froze in shock. Almost 30 years ago, this move was revolutionary and went down in gaming history as one of the most innovative and iconic moments. The game was the brainchild of one of the most popular video game minds of our time, Hideo Kojima, a Japanese designer who has been idolized for his design and storytelling for years now. It was the first time a game had ever attempted something like this. It's creepy because it's supposed to be a genuinely unsettling experience. And this was celebrated as a creative, deliberate little privacy breach that enhance the experience of the game. Cut to about 30 years later, and Microsoft introduced a co-pilot gaming assistant to the Xbox mobile app.

It could track your play history. It could even summarize your last session and read your screen to coach you into what you need to do next if you get stuck. It could pretty much do what Psychomantus did and plus it was supposed to help, not just scare. But this time, there was no celebration. Everyone hated it. In March 26, a clip of Microsoft's own demo of co-pilot guiding a play-a-through game called Sea of Thieves went viral.

In a bad way. Even Xbox fans who are generally quite passionate defenders of the platform were repulsed. Gamers didn't like the feeling of being watched. It felt intrusive, pointless. and a little bit like being talked down to. Kinshuk Sunil, a developer with more than 20 years of experience I spoke to, said this.

An AI bot is sort of sitting with you and helping you navigate the game, make choices on your behalf, right? Like, I'm not sure if that's a good example, but it's like you go watch a foreign language movie in cinema hall and you have somebody sitting right next to you, translating it for you in the run-bound. But then in early May, just two months later,

everything changed. Asha Sharma, who had just stepped into the road, as the new CEO of Xbox in February this year, killed co-pilot via a tweet on X. This came as a shock to most, because just in March, when the Sea of Thieves clip went viral, co-pilot was all set to appear on the console as well.

Not just that. In 2025 alone, Microsoft had spent over $60 million in televised advertising for co-pilot, pushing it aggressively on all their products. And Xbox should have been an easy, play for co-pilot. The gaming community is tech-savvy. Gaming is one of the only entertainment formats where you get stuck and it's a market where people spend a lot of money to improve their experience. Something like co-pilot should have been the perfect fit over here. So why did it die so fast?

And so publicly. Welcome to Daybreak, a business podcast from the Ken. I'm your host, Richard Berguise. And every day of the week, my co-host, Nikash Ram and I will bring you one news story that is worth understanding and worth your time. today's Wednesday, the 27th of May. Now, here's the irony in Copilot's very sudden death on Xbox. Barely a month before Xbox killed Copilot, Microsoft had been called out by users for describing Copilot in its own terms of use as for entertainment purposes only, warning that it could make mistakes and advising users to use Copilot at their own risk

and not rely on it for anything important. considering that Xbox is arguably Microsoft's only entertainment product, co-pilot being laid to rest over there so suddenly made even less sense. When the feature was first introduced, the logic was this. Gaming was one of the only leisure activities where you could get stuck. Considering how frustrating that experience can be, it made some sense to reduce that friction by introducing a bot that could prompt or coach you the player out of a bump. But here's the thing.

As Nitin Govind, a game developer and UIUX designer explained to me, friction is actually an integral part of the gaming experience. He told me about the psychological concept called a flow state, which was recognized and named by Hungarian-American psychologist Mihai Chixent Mihai. It's a principle that game design follows, maintaining a careful balance between challenge and ease so that players stay in a focused, engaged mental zone. Too much friction can cause a player to get frustrated and quit.

Too little friction and they won't feel challenged, they get bored and then they quit. I also spoke to a game designer who works in the industry and they believe that an assistant-like co-pilot ends up smoothing over the good kind of friction and interrupts the flow state. A Google study from 2024 showed that 36% of Gen Z players quit games specifically because of disruptive ads or intrusive features and not because of difficulty, price or content. The designer explained that players want agency and they want to feel like they've mastered the necessary skills.

But if AI starts playing for them, it quickly erodes the satisfaction of playing itself. So Xbox's first failure was a fundamental technical misunderstanding of game flow and game design and not really knowing what problem co-pilot was really trying to solve. The other reason is a bit more philosophical. Kinshuk Sunil, the developer you heard earlier, is also the co-founder of a 17-year-old game dev community in India,

which is almost 10,000 members strong, and he's also the co-founder of a gaming studio called Second Quest. Here's what he said about what's important to gamers. The whole game slop and community player backlash, against AI originates from the fact that like any medium of art, games also is a very highly creator-focused activity. When you play a Kojima game versus an American game, like the fact that somebody has built this experience is the most identifying field element of the industry.

The player backlash also sort of comes from whether the creative energies of the developers were completely utilised or not. Especially when it comes to narrative or puzzle-based games, there's an unspoken contract between the player and creator. And with a feature like the gaming assistant, that contract was being interrupted at every point. Like what Sunil said earlier about having someone translate a movie to you in real time. It's annoying. And you see, the movie analogy he uses. makes even more sense because good game design is cinematic. And Sunil's mention of Kojima also adds to it.

Brady Ng, the deputy editor at the Ken and also one of the co-hosts at Zero Shot, who games occasionally described Kojima as gaming's equivalent of some of the greatest film directors out there. He said the intricacy of Kojima's words make players feel like they've lived a life by the time they've finished. That's the level of emotional and cognitive engagement. gaming demands. You're not just a passive viewer. You're actively making choices and shaping the story. You are the main character. And that explains why it left so many people so frustrated. Like I mentioned earlier, some people even found the integration to be something of an insult. Here's what Luke Plunkett and Enraged Reviewer writes for a website called Aftermath.

Is this how little these people think of us? Do they think we, adults spending what little free time and money we have on these specific video games are simply enormous babies that we are mindless meat sacks happy to just sit there mashing on an AI button as content is shoved down our throats. Like what the fuck is this feature for? What do they think we are playing video games for?

Video games are many things. They are art, they are challenged, they are entertainment, they are product. And the angle towards co-pilot also goes just beyond the artistic. As with every AI critique, data privacy was also obviously an issue.

But I think the other important aspect was also that in a way the co-pilot model was learning from player actions. Right? That is how it would help solve the game, right? So there was also this concern about

the kind of data it is collecting and how it will be used. Basically, players had also discovered that co-pilot was scanning gameplay screenshots and sending data to Microsoft servers. A setting that allowed model training via text was also enabled by default without any clear consent from users as well. Things only got worse when a Microsoft spokesperson actually clarified that co-pilot was simply pulling from public sources via Bing.

And then in the same statement, advice that players verify its answers against other trusted sources. Reviewer Thomas Wilde at Geekwire put the response the best. If you have to fact-check it anyway, what exactly is it for? He was especially offended that the bot was taking away necessary web traffic from actual humans who were already sharing the guidance. And that ties into the community aspect. You see, the assistant was pitched as something that would maintain a player's immersive experience because now they would not have to go looking for a guide or walk through and leave the game.

But for a community that's pretty much its own ecosystem, that doesn't feel exactly right. Now I am not a habitual gamer by any means, but I have dabbled a couple times. When I got stuck in Batman Arkham night, I spent hours combing through Reddit threads to find the most recommended guides and walkthroughs. I came across so many comments and conversations then.

People were asking follow-up questions, praising mechanics, complaining about something or just being excited about the game. And none of this ruined my emotion in the game at all. If anything, it only deepened. Knowing why so many people love a game and will choose to struggle through it is better encouragement than a bot could ever be. At least the stage AI is currently at.

So when I asked Sunil if Xbox had fundamentally misunderstood who it was selling to, here's what he said. Yeah, so I think because the AI landscape is changing so fast, I don't think even the larger corporations are getting sufficient. time to make informed decisions. I think that's the primary challenge. Co-pilot is a different example, but look at what happened with

Nvidia's DLSS-5. So, Nvidia launched the new algorithm for DLSS, it's a technique used to render higher quality frames without being very taxing to your machine, right? So the earlier

versions, like till DLSS 4, this was an AI-trained model that would upscale your images. So even for example, if you're playing the game at 2K, it is actually rendering everything at full HD and then using AI algorithms to upscale it to 2K so that you see a rich quality gameplay. With DLSS5, what they did was they took that upscaling to a level where the in-game characters and elements started looking very lifelike, right?

As if it's like a full motion video instead of a game. So from a technology perspective, I think it's phenomenal that something like that can be done. But from a player and playing experience perspective, it just takes a lot of value away. It's such a clear example of how a corporation's poor understanding of its consumers actually came around to bite it. To Xbox's credit, the pivot seems to have garnered goodwill already. And Sharma's statement does promise a focus on community. But where can the company really? He direct its AI efforts too.

As far as gaming assistants are concerned, the general understanding is that there is no version where it works. Sunil and Netin both stress that hints and skill development in games is something that's baked into the design itself. A good game will give you the skills necessary to beat it progressively so that the player's ability increases with the difficulty of the game. The entire thrill in playing a game comes from facing a challenge, figuring out a solution and then moving forward.

like getting better improving and then moving forward. And if you're using AI bots to help you do that, then there is no point playing that game in the first place. Considering what the gaming community's reaction to co-pilot looked like, it might seem like it's a little resistant to change. But Sunil disagreed. No, I wouldn't say that the gaming community has a resistance to change. In fact, if you look at the history of gaming,

it's adoption of technology disruption. I think gaming has been very accepting of changes and disruption. I don't think there's any resistance to adapting it. I think the resistance that we see is more, like it eventually goes back to the same set of concerns. The concern about training data, the concern about the human element, concern about taking over jobs, things like that.

The only thing I would say about the gaming experience is the whole quality aspect of the game slot that's happening. I don't agree with everything. thing will eventually become game slot because, you know, at the end of the day, AI is a tool. There will be a lot of creators who will use it very wisely as well and create rich experiences. So not everything would be game slot. But yeah, the barrier to entry for making a game. I'm not saying a good game.

I'm just saying a game. A game that will be released. That barrier to entry has been reduced, right? So the inflow of slop would be a lot. but there will always be a sufficiently large set of creators who will be able to use this technology and the technologies to come to create high-quality content. So what could help is if AI were very intentionally,

very artistically part of the game. Kind of like how Kojima made that creepy little trick work with Psychomantis. On the developer's side, AI is already used to generate assets, catch, bugs, and handle repetitive coding tasks. The general consensus amongst everyone I spoke to is that it's great for the manual tasks, which then frees developers up to do the human stuff that matters, like ideation, storyline, narrative.

That being said, Sonal explained that using AI in developing NPCs or non-playable character interactions in games have been a very obvious case of a successful player-facing integration. almost 100x of non-playing characters. That can be very easily replaced with generative AI content where you pass on the current state attributes and you say, okay, this is what is happened in the world,

this is the context to this character, this is the personality of the character. Using that, generative AI can create a lot of interesting narrative outcomes for the players to deal with. Let me give you an example that breaks this down. Now, to be clear, this example is from before the Gen AI boom of 2023. But the principle stays the same and shows how an in-game AI system could work really well. Shadow of Mordor, a game based on the Lord of the Rings series, had built in something called a nemesis system.

Basically, all the villains and bad guys in the game would remember you as a player, evolve rivalries and even grow stronger from past encounters. No one found this to be intrusive. In fact, this just enhanced the game and its immersion. Another way is to make additional AI-generated characters that can take actions on your behalf. Ubisoft, a gaming studio, announced a prototype last November called teammates, a first-person shooter where your AI squadmates respond to real voice commands like take cover or focus fire. The game is still in very early development, but the demos have already created again. excitement. And reviewers have praised the fact that the AI still operated within the very clear

rules set by the developers. And Xbox already has some plans to plug in AI in more meaningful places. Sharma's tweet while killing copilot as a chatbot clarified that AI wasn't really going away. She wrote on X, we are refocusing our AI efforts to solving player problems like enhancing real-time graphics, improving discovery and deepening personalization. AutoSR is a good early example. Models reconstruct high-frequency detail from lower rest frames. Remember how Sunil explained what Nvidia's DLSS-5 did? It made games look so realistic.

It triggered a kind of uncanny valley response for gamers. Things didn't look the way they were supposed to, so they got uncomfortable. But that being said, the tech is still exciting. And the auto-sr, or automatic super-resolution that Sharma mentions, is supposed to do the same thing. but at a less aggressive scale. Improved graphic quality without creating lag and without giving players the creeps. Largely speaking, the goal overall hasn't really changed.

When the co-pilot gaming assistant was introduced, the point was to reduce friction. Instead, it ended up introducing surveillance friction and performance friction. The new goal now is to still reduce that friction. But this time, it looks like Xbox may have figured out the right place to do it.