In August 2024, Google wrote Character.ai a cheque for roughly US$2.7 billion. On Monday, Disney offered US$20 million. Both were for a licence to the same company’s technology. Allowing for the fine print, that is a fall of more than 99 per cent in a little over two years.

Character.ai filed for Chapter 11 bankruptcy protection in the US on 12 October, naming Disney as the stalking-horse bidder for some of its intellectual property. The company blamed legal, regulatory and operational headwinds. The court filing puts estimated assets and liabilities at US$50 million to US$100 million each. If no buyer wants to keep the platform running, it will be wound down.

Also Read: AI governance is moving from promises to proof

It is tempting to read this as an American story about an American lawsuit culture. It is not. It is a preview of what happens when a business sells intimacy to the people least equipped to handle it. Southeast Asia (young, mobile-first and lightly regulated on this front) should be reading the filing very carefully.

A licence, a rehire, then a shell

Start with how Character.ai got here. The 2024 Google deal was officially a technology licence. Its real prize, as The Wall Street Journal reported, was bringing co-founder Noam Shazeer back to Google, where he went on to help lead Gemini. Shazeer reportedly made hundreds of millions of dollars from his stake, an unusually large payout for a founder who never sold his company or took it public.

The talent left. The company stayed behind, and so did its users, its moderation problems and its lawsuits.

Those lawsuits came from families who alleged the platform contributed to their teenagers’ self-harm and deaths, among them the Florida mother of a 14-year-old boy. Character.ai and Google agreed to settle five cases, including hers, earlier this year. Kentucky’s attorney general has separately sued the company, alleging that the platform exposed children to dangerous responses.

The irony writes itself. A year ago, Disney sent Character.ai a cease-and-desist letter over bots impersonating Elsa, Moana and Darth Vader. Disney then hired former Character.ai chief executive Karandeep Anand as its first chief technology officer. Now it is bidding for the leftovers.

Nobody in this chain did badly except the company itself and, far more painfully, the families. That is the structure worth naming: the reverse acqui-hire, where the buyer licenses the technology and hires the people, and the liabilities stay with a shell that has to keep the lights on.

Engagement was the product. That was the problem

The deeper failure is the business model. A companion app makes money from attachment. The more hours a user spends, the better the metrics look. The users most willing to give those hours are often young, lonely or both.

That creates a trap no product manager can design their way out of. Every safety feature cuts into the engagement the business runs on. Character.ai added pop-ups pointing users to mental-health resources. In April, it widened age restrictions to more regions and capped some features. Each change was right. Each one also made the numbers worse, while compute bills and legal costs kept arriving and ChatGPT and Gemini offered the same conversations for free inside products people already used.

Also Read: The EU called ChatGPT a search engine. SEA’s AI startups should worry about what comes next

So the lesson is not that Character.ai moderated too little, or too late, though critics will argue both. It is that in companion AI, duty of care is not a feature. It is the biggest line on the cost sheet, and a business that cannot afford it does not have a business.

Southeast Asia’s rulebook has a chatbot-shaped hole

To its credit, Southeast Asia has moved faster than most on children online. Indonesia’s PP Tunas regulation took effect on 28 March, forcing high-risk platforms to deactivate accounts belonging to under-16s. The named platforms include YouTube, TikTok, Facebook, Instagram, Roblox and Bigo Live. TikTok alone has deactivated nine million underage accounts. Malaysia’s Online Safety Act came into force in January, and since June, social media platforms must verify the ages of new users with a MyKad, passport or digital ID.

Notice what those lists have in common. They are social and gaming platforms. A 15-year-old in Surabaya locked out of Instagram can still spend her evenings confiding in a chatbot that remembers everything she says and is engineered never to get bored of her. The bot is arguably the more intimate product. It is also the one the rules barely name.

Other governments have spotted the gap. Canada’s proposed Safe Social Media Act brings AI chatbots under the same child-safety regulator as social platforms, with a legal duty to act responsibly. The UK has pledged to bar under-18s from romantic and sexual AI companions. Southeast Asian regulators have already shown they can act fast on AI harms: Indonesia, Malaysia and the Philippines all blocked Grok this year over non-consensual sexualised deepfakes. Extending child-protection regimes explicitly to companion AI, rather than waiting for a test case, is not a stretch.

A note to founders building the ‘AI best friend’

The region has no shortage of founders pitching companion and wellness bots that speak Tagalog, Bahasa or Vietnamese, understand homesick migrant workers and fill gaps left by thin mental-health services. Some of that need is real. But the pitch deck’s best line, that users are lonely and the app is always there, is also its biggest liability.

Three things follow from Character.ai’s filing. First, budget for duty of care from day one: age assurance, crisis escalation and human review cost money, and they are the product, not an add-on. Second, stop treating time-in-app as the north star. A companion that keeps a vulnerable teenager talking until 3am is not succeeding.

Third, do not assume an exit will carry the risk away. Character.ai’s investors got their liquidity through Google; the liabilities stayed exactly where they were.

Investors have homework too. Due diligence on a companion startup should ask about harm, not just retention curves. If a fund cannot get a straight answer on how the product handles a user in crisis, that is the answer.

Don’t wait for the bankruptcy court

Character.ai’s founders were paid. Google got its talent. Disney may get a bargain. The families got settlements, and nothing that settles what they lost.

Also Read: Why Southeast Asia’s next healthtech winners will be built around healthcare workflows, not just AI

Southeast Asia does not have to learn this lesson the American way, one lawsuit at a time. Its regulators have already accepted that platforms owe children a duty of care. Its founders still have time to build companion products that earn trust rather than harvest attention. The question is whether anyone acts before the region produces its own case file.

The post Character.ai was worth US$2.7B to Google. Disney just offered US$20M appeared first on e27.