Root/Wen An AI (artificial intelligence) company with approximately 12 million monthly active users has still reached the stage of bankruptcy and restructuring.

On October 12th, the American AI companion company Character, with a valuation of over $1 billion at one pointAI applies for bankruptcy protection and seeks to sell assets. According to the bankruptcy documents, in August 2026, the company's revenue was approximately $5.3 million, but its operating expenses amounted to $13.3 million, which has not been included in legal fees. Long term income deficit is an important aspect of its business difficulties.

For Internet enterprises, user size and usage duration have always been important indicators to measure the commercial potential of products. In 2023, venture capital firm a16z will lead the investment in CharacterWhen AI raised $150 million in funding, it specifically mentioned that its users use it for an average of about two hours per day. However, the user stickiness that investors once valued did not give CharacterAI brings enough revenue to cover expenses.

Unlike traditional content platforms, every conversation accompanied by AI requires generating new responses, and the longer users chat, the higher the computational cost. In the first eight months of 2026, Character. AI's computing power expenditure alone reached 56.1 million US dollars; By August, its operating expenses were still about 2.5 times its revenue.

However, the cost of computing power is not enough to explain the CharacterWhy did AI go bankrupt. According to bankruptcy documents, the company has had contact with about 28 potential investors and buyers, but the litigation and regulatory risks surrounding the safety of minors have deterred many trading parties. In addition, due to changes such as the loss of core technical talents, this once favored company by capital ultimately chose to seek a way out through bankruptcy proceedings.

And the emotional dependence issues involved in these lawsuits are also related to the way AI accompanies and attracts users. In CharacterOn AI, people can create their own AI "friends" and "lovers" to chat with for a long time.

The platform also introduced a feature that allows characters to actively send messages. A person who seems willing to listen and respond to oneself at any time naturally creates a sense of closeness.

Such services certainly have their value. A study published in the Journal of Consumer Research in 2025 found that AI companionship can temporarily alleviate feelings of loneliness in specific contexts. But when emotional connections become a part of product appeal, companies need to be particularly cautious in how they manage these relationships.

Making users reluctant to leave may be the hidden danger and temptation of this business. Enterprises certainly hope that users will use it frequently and invest more time and emotions. But if we constantly cater to or even reinforce users' unhealthy dependence on virtual relationships in order to strive for longer usage time, we may cross the security boundary.

The lawsuits faced by AI have provided a warning. In 2024, after a 14-year-old American teenager committed suicide, his mother sued CharacterAI， Accusing his son of developing an unhealthy emotional dependence on chatbots and believing that the related interactions had an impact on the tragedy. In January of this year, Character. AI and Google have agreed in principle to settle five lawsuits involving minors.

In October 2025, Character. AI announced the cessation of providing open AI chat services to minors. However, the company did not immediately interrupt the relevant conversations, but gradually shortened the chat duration and collaborated with external organizations to provide emotional support and exit arrangements for underage users, guiding them to find real-life support resources.

Two years ago, an average daily usage time of two hours was still an advantage that investors valued; Two years later, the company needs to arrange support measures for underage users to gradually withdraw from open chat. This change is quite intriguing.

AI accompanies human emotional needs in business operations. How to make users stop in a timely manner and prevent unhealthy dependence should be a part of product design, rather than waiting for security disputes to occur before making remedies. In September 2025, the Federal Trade Commission of the United States will include CharacterSeven companies, including AI, have issued investigation orders requiring them to submit measures to ensure the safety of young people. In July of this year, China implemented the "Interim Measures for the Management of Artificial Intelligence Personification Interactive Services", which also made provisions for inducing addiction and dependence, risk reminders, and convenient exit.

If companionship services rely on deepening unhealthy dependencies to sustain growth, even if they ultimately achieve profitability, it is not a good business. As a business, AI companionship requires' knowledge to stop '. This restraint reflects the company's reverence for business ethics and also reflects its values. Without the correct values, it is easy to blur the boundary between compliance and violation, ultimately devouring the foundation of a company's existence.

(Author's Commentary)