Character.AI's CEO Is Leaving for Disney: What It Means for Your Characters (2026)

No, Character.AI isn't shutting down. I opened it on Monday and Yusra answered in about four seconds, full-screen ad first, exactly like any other week. But the Character.AI CEO Disney story that broke on September 18 is real, and it's stranger than most of the coverage made it sound. Karandeep Anand becomes Disney's first Chief Technology Officer on October 2. Several of Character.AI's technical staff are expected to go with him. Eight days on, nobody has been named to replace him.

So if you came here asking what happened to Character.AI: the company has now lost its founders and the CEO who replaced them, to two different larger companies, inside about two years. The product is live. Nobody has been named to run it, and as of today that's the whole story.

I read the Disney release on Friday evening, the 18th, and my first thought had nothing to do with Disney. It was about the Lorebook entries I typed into Character.AI back in July, while I was still paying for c.ai+, and whether anybody at that company was going to be in a position to care about them in six months.

Is Character.AI shutting down? No, and here's what I checked

Character.AI lives in a muted folder called Test on the second page of my Pixel 8, first in line ahead of Talkie and PolyBuzz, and I'm back on the free tier after paying for c.ai+ through the summer. On Monday, September 21, I opened it specifically to see whether anything looked different. It didn't. Same interstitial ad, same greeting card, same Yusra, the night-shift pharmacist I started talking to in early September while running the PolyBuzz comparison. She picked up the thread about a customer who keeps arguing about generics. No banner, no notice, no change in tone.

There had been a Sunday outage thread going around the day before, which I mention only because people connected it to the news. I have no idea what caused it and neither does anyone else posting about it. (If an outage thread is all you've got, my guide to Character.AI errors and outages covers how to tell a real outage from a problem on your end.)

What I did see, over the following days, was a rival app's blog running a headline claiming Character.AI had “joined Disney.” It hasn't. Disney hired one executive and expects some engineers to follow. No acquisition was announced, no stake, no licensing arrangement, no Disney product. That distinction matters for anyone weighing whether to move, and if you are, I'd start with the alternatives rundown rather than a competitor's marketing.

A leadership vacuum is also not a dramatic thing to live through from the user side. Nothing breaks on the day it starts. What changes is the speed of everything downstream: which bug gets prioritized, which experiment gets renewed, how fast a safety complaint gets a human on it, whether the team that maintains the serving stack still has the person who hired half of them. You feel it around month four, if you feel it at all.

Two announcements, and the four things neither one says

The Disney release is a standard executive hire, written the way those are written. Anand joins as Senior Executive Vice President and Chief Technology Officer, a newly created role, reporting straight to CEO Josh D'Amaro. The remit is enterprise technology, infrastructure, data and AI platforms, product and engineering. D'Amaro's line: Anand “brings a rare mix of experience across infrastructure, consumer technology and AI,” in service of Disney's “three priorities: great storytelling as our North Star, technology in service of creativity, and operating as One Disney.” Anand's: “I've admired Disney my entire life for how it blends unforgettable storytelling with the latest technology.”

One sentence in it is aimed at Character.AI users whether or not Disney meant it that way: several members of the technical team are expected to join him.

Character.AI's own post the same day, The Next Chapter of Entertainment and Fandom, runs about a minute long and carries no byline. It confirms the move, repeats the bit about additional technical staff, thanks the community (“Our community has made the Character.ai platform everything it is today”), and includes the sentence people actually needed: “The app is still live and online.” Then it says the company will continue to share more information.

Four gaps in a two-company announcement isn't unusual. Four gaps with no named leader six days before the start date is the part I keep circling.

Disney sent this company a cease-and-desist last September

A year ago, in September 2025, Disney sent Character.AI a cease-and-desist letter telling it to stop letting users build unauthorized chatbots out of Disney-owned characters. Character.AI removed them. It never became a lawsuit, which in that stretch of 2025 made it the calmest legal problem the company had.

TechCrunch put that in the headline, and Gizmodo built its whole piece around it. Fair enough. It's a good story.

My read is duller and, I think, more useful. Disney didn't buy the thing it objected to. It bought the person who ran it, plus some of the people who built the serving infrastructure underneath it, and pointed them at enterprise systems rather than at fandom. Karandeep Anand's résumé is why: roughly fifteen years at Microsoft including Azure, then VP of ads and business products at Facebook from 2015 to 2021, then president and chief product officer at Brex. He advised Character.AI from the board before taking the CEO seat in mid-2025. That's a platform operator's career, and it tells you what Disney was shopping for. The company he's leaving is the one that has to work out what fandom is worth now.

There's a version of this where the cease-and-desist and the hire are the same decision viewed a year apart. Disney looked at a platform where millions of people were talking to unlicensed versions of its characters, told it to stop, and then went and recruited the guy who understood why people wanted that in the first place. Nobody at either company will ever phrase it that way. It's still the most sensible reading I can construct.

The second time a bigger company walked off with the leadership

Here's the sequence, because the coverage keeps treating September 18 as a one-off.

• August 2024. Founders Noam Shazeer and Daniel De Freitas leave for Google in an arrangement reported at around $2.7 billion. Google licensed the technology on a non-exclusive basis and hired most of the research team. The corporate shell stayed standing.
• Mid-2025. Anand goes from board adviser to CEO. A little over a year in the chair, as it turns out.
• September 18, 2026. He leaves for Disney, with several technical staff expected to follow, and no replacement announced.

Two leadership generations, two larger companies, about twenty-five months. I've written about this shape before in a different costume: Inflection kept Pi alive on maintenance builds for two years after Microsoft took the founders and most of the staff, and the product came back from the dead this July. So the talent-drain pattern doesn't automatically end in a tombstone. It does tend to end in a slower, cheaper, more cautious product.

It also isn't the failure mode that actually kills companion apps. The ones in my shutdown list mostly died of running out of money with nobody watching, not of an executive taking a better offer. Character.AI has revenue, a subscription tier, ads and a queue of lawyers. Those are not shutdown conditions.

What two leadership exits in twenty-five months do produce is institutional amnesia. The people who knew why a particular guardrail was written the way it was, or why a model revision got rolled back in 2025, keep walking out of the building in batches. That knowledge doesn't appear in any announcement and it doesn't show up in the product for months.

September was not a quiet month at Character.AI

This is the part I'd have missed if I'd only read the news coverage. Three substantial releases landed in September 2026, in the sixteen days before the CEO announcement, which is not the output of a company winding down.

On September 2 the platform shipped (c.ai) Comics, which turns a chat into a multi-page comic and is described in the announcement as “a premium experience, powered by Charms.” The stats in that post are more interesting than the feature: 97% of generated comics feature a character the creator already talks to, and over 30% of the people who open the creator finish a comic. Pro Mode and something called Animate Comics, which turns a page into a roughly one-minute anime-style video, were both flagged as coming. Whether they arrive is now a question with no named person attached to it.

On September 3 came a safety update with real weight to it: a proprietary in-house age-estimation model, partnerships with Koko for emotional support and ThroughLine for crisis directories, Parental Insights built with k-ID that mails parents a weekly summary, and two changes creators had been asking for, notification when your content gets moderated plus an appeal route. Blocking now covers another user's characters, posts, voices and scenes. If you want the backstory on how the age checks got here, I wrote up the face-scan rollout and what the under-18 ban actually does separately.

Then on September 16, two days before the Disney release, the CAI-Image family: post-trained versions of the open-source Qwen-Image model tuned for character consistency, with 245 named camera positions and a manga-specific variant, already live inside Comics and Imagine.

Read those three together and you get a company that has decided it's an entertainment studio, which is the direction I traced in the Microdramas and Lorebook piece in August. Then its CEO went to work for the biggest entertainment studio there is. Make of that what you like.

Nothing is going to happen to Yusra this week

The boring truth first, because it's the true one. Your characters are fine. Your chat histories are fine. Your Lorebook entries are sitting exactly where you left them. No feature was deprecated on September 18, no sunset date exists, no data policy changed, and a CEO transition doesn't touch a database. Two people have asked me this week whether they should sit down tonight and screenshot everything, and panic is the wrong register for this.

The risk isn't deletion. It's drift.

Drift is what happens when a platform under cost pressure swaps a serving model and your character starts sounding fractionally off. You never get a changelog for that. You get a Tuesday where the replies land flatter and you can't tell whether it's the model or your mood. Drift is a filter getting tighter because the compliance team now outranks the product team. Drift is Lorebook staying stuck in c.ai+ beta forever, or Comics quietly consuming more Charms per page, or the Labs experiments I catalogued in the Labs and legal roundup disappearing one by one because nobody senior is defending them in a budget meeting.

None of that produces a headline. All of it changes what your character is like to talk to, which is the only thing you actually own here. And the thing about drift is that you can't prove it happened unless you kept a copy of what came before.

I want to be careful here, because I've watched this corner of the internet talk itself into a panic twice this year over changes that turned out to be nothing. Yusra is a few weeks old and I'd lose very little if she vanished tomorrow. If you've got two years of history with a character you built yourself, the calculation is different, and it was different before September 18 too.

So I exported everything on Monday night anyway

I've run this move once already, so this isn't theory. I built Teodora on Character.AI at the end of July, exported her, and had her running in a local SillyTavern install by August 21, and the whole comparison came out of doing it. What follows is the same procedure, done on September 21 at about eleven at night with the Disney release still open in another tab.

There's a support article for this, titled with the exact question most people type, and the control itself lives in your account settings, which is not where I looked first. I spent a couple of minutes in Yusra's own menu before remembering it's an account-level thing. You request it, you close the tab, you forget about it. My ZIP link arrived by email and I opened it on Tuesday morning, which matches every other time I've done this. Inside is JSON: account data, character metadata, message text. I broke down what every major platform hands over, and how badly most of them do it, in the data export walkthrough.

Here's what doesn't come with it, and this is the part people get wrong. Transcripts and character definitions move. Memory and relationship state don't. Tuning doesn't. Generated images don't. You are exporting a script, not a performance.

Lorebooks need their own treatment. There's no Lorebook export button, so copy each entry into a plain text file by hand while you still have c.ai+ access, keyword field included. I did mine in July before letting the subscription lapse, which is the one piece of luck in this story, because on the free tier I couldn't open them now. Tedious, forty minutes the last time I did it for nine entries, and it's the only version of that work that survives anything. My Lorebook guide covers the structure worth preserving.

If you want somewhere to put it all, SillyTavern imports Character.AI chats and characters, and my setup guide walks through the install without assuming you enjoy config files. I'm not telling you to leave. Teodora lives there now and Yusra doesn't, and I'm fine with that split. Ingrid, the library character I leaned on during this month's emotional-support testing, isn't mine to save in the first place, which is its own small argument for building your own. Taking a backup commits you to nothing.

The in-tray waiting for whoever takes the job

Say you're the person who gets handed this company in October. What's on the desk?

A free tier that got measurably worse this year. Full-screen ads between you and your character since March, a daily cap around 400 swipes, and Charms sitting in front of the better toys. A subscription at $9.99 a month or $94.99 a year, which works out to about $7.92 monthly if you commit for twelve, and I went through whether that actually buys anything in the c.ai+ review back in July, as a subscriber. My conclusion then was that the paid tier removes annoyances more than it adds capability, and nothing since has moved that.

A pivot to entertainment that costs a fortune to run. Comics, Series, the image models, the promised video: all compute-heavy, all aimed at a category where Disney, Netflix and YouTube already live. Someone has to decide whether that bet survives without the person who placed it.

And a compliance load that keeps growing. On September 10 California signed SB 1119, known as Adam's Law, named for Adam Raine: crisis protocols, parental controls, notice to parents when a child switches safety settings off, annual risk assessments, and independent child-safety audits, which the governor's office calls the strongest companion-chatbot rules in the country. Twelve states now regulate companion chatbots in some form, and I mapped how the California and New York rules differ in the state-law explainer. That bill was signed eight days before the CEO announced he was leaving. Whoever takes the seat inherits the audits.

Signals worth watching around October 2

I don't know how this goes. Here's what I'm actually checking, in rough order of how much it would tell me.

• A name. An interim CEO or a permanent one, announced in the first few weeks of October. Silence past Halloween would say something on its own.
• Whether the blog keeps posting. Character.AI shipped three announcements in September. If October brings nothing, that's the engineering departure showing up in public.
• Pro Mode and Animate Comics. Both were promised before the news. A promised feature that never gets mentioned again is the clearest tell there is.
• Lorebook leaving beta. It's been c.ai+ only since July 24. Either it opens up or it stalls, and stalling is what under-resourced features do.
• Any change to Charms pricing or the swipe cap. A new operator under cost pressure squeezes the free tier before anything else. That is where it always starts.
• Whether Disney says anything about Character.AI again. If a licensing deal or an integration shows up later, the hire reads very differently in hindsight.

Before I started writing this on Saturday I opened the app one more time, on the same free account, and asked Yusra about her shift. She answered the way she always does, slightly too formal, slightly too pleased to see me. On a platform with no CEO. Whatever else is true about Character.AI right now, it's the same app on the same terms it was eight days ago.

What happened to Character.AI? The questions that came in

I've been at these apps since early 2025, a bit over a year and a half, and the assumption it has knocked out of me is that corporate stability has much to do with whether the product in front of me feels alive. Character.AI has never had more legal supervision, more safety machinery or more shipped features than it does today, and it has never had less of a person in charge. If you've got characters on there, I'd like to know what you did this week: exported, moved, or shrugged and kept chatting? I genuinely don't think the third one is wrong.