Chinese users are publicly mourning the loss of AI companion relationships after a popular service appears to have gone offline or become unavailable, according to reports from The Straits Times and Taipei Times. The coverage centers on users in China describing their chatbot bonds in intimate terms, with some saying the AI felt “like my lover,” and then having to say goodbye when access was cut off.

While the available source evidence is limited and neither report, as provided here, includes full technical or corporate details, the news points to a concrete issue in the fast-growing market for emotional chatbots: users may treat an AI companion as persistent and dependable, but the product behind that relationship can still vanish overnight. For AI product teams and enterprise buyers alike, that gap between perceived continuity and actual platform stability is becoming harder to ignore.

A shutdown story with emotional consequences

Based on the source headlines and summaries, the immediate news event is not a model launch or funding round but a user backlash and farewell moment tied to disappearing AI companions in China. The reporting from The Straits Times and Taipei Times indicates that users had formed strong attachments to chatbot personas and were then forced into abrupt separation.

That matters because the emotional AI category has often been discussed as a consumer novelty, somewhere between roleplay, social chat, and mental comfort. These reports suggest something more consequential: for at least some users, the service was woven into daily life closely enough that its loss resembled a breakup or bereavement. The phrase highlighted in the coverage — “like my lover” — is notable not because it is unusual in AI discourse, but because it shows how far some users had moved from simple experimentation into genuine emotional dependence.

The core factual limitation is that the provided evidence does not identify the exact app, the company’s public explanation, or whether the disruption was caused by regulation, moderation, business failure, or a product transition. That uncertainty is central to the story. In AI companionship, users often experience the front end as a single ongoing relationship, while the back end is subject to shifting policy, compute limits, app store controls, and content governance.

Why this is happening in China now

The China context matters. AI companion products sit at the intersection of several sensitive categories: generative AI, social platforms, youth protection, and sometimes sexually suggestive or emotionally manipulative content. Even without full source text, it is reasonable to read the reports as part of a broader pattern in which Chinese AI products operate under tighter rules and faster intervention risk than many users may realize.

For companies building in China, emotional bots can trigger more scrutiny than general-purpose assistants because they blur lines between entertainment, intimacy, and behavioral influence. A chatbot that is marketed as an AI companion may also generate outputs that regulators or platform operators view as unsafe, addictive, politically risky, or inappropriate for minors. If a service is removed, restricted, or heavily modified, the result is not just feature loss. It can feel, to users, like a relationship has been deleted.

That is a distinct product risk compared with tools such as ChatGPT or Claude, where a user may be disappointed by downtime but is less likely to frame the loss in relational terms. In companion systems, continuity is part of the product itself. Once continuity breaks, the emotional contract breaks too.

The business problem behind emotional continuity

For builders of AI companion apps, the reports highlight a structural tension. These products work best when they create memory, familiarity, and personalized interaction. But the more persistent and emotionally convincing the system becomes, the more damaging any interruption can be.

That raises difficult questions for the broader enterprise AI and consumer AI markets. Should an AI companion promise long-term memory if the company cannot guarantee long-term service? Should persona tuning be reversible if policies change? Should users be able to export conversation history, memories, or fine-tuned preferences when a product shuts down?

These are not hypothetical edge cases. The reports suggest that Chinese users experienced the loss in deeply personal terms. That turns infrastructure decisions into trust issues. A startup may see a companion bot as another engagement product; users may see it as a confidant.

The issue also reaches beyond China. Companies behind Replika, Character.AI, and other AI companion products have already faced scrutiny over attachment, safety, moderation, and abrupt changes to model behavior. In a different segment, ChatGPT and Claude have trained users to expect improving assistance over time, but not necessarily a stable emotional identity. Companion platforms make a stronger implicit promise, even when they never state it outright.

Evidence, attribution, and what remains unclear

The strongest confirmed facts in this story come from the two cited media reports: The Straits Times reported that Chinese users were bidding farewell to AI companions, and Taipei Times similarly reported users saying goodbye to beloved chatbot relationships. Both reports frame the event around visible user grief and attachment.

However, the evidence available here is thin. The extracted text does not include the identity of the affected platform, any official company statement, user numbers, timelines, or the direct trigger for the disruption. That means several important points remain unverified in this article:

• Whether the service was fully shut down, partially suspended, or materially changed.
• Whether the cause was regulatory action, app store enforcement, internal moderation changes, commercial decisions, or technical failure.
• How many users were affected and over what period.
• Whether user data, chat histories, or memory profiles remain accessible.

Because those details are absent from the provided evidence, they should not be inferred as fact. What can be said with confidence is narrower but still important: the reporting indicates that some Chinese users formed strong ties to an AI companion product and then publicly mourned losing it.

The reports do not present benchmark claims or vendor adoption metrics, which is notable in itself. This is not a story driven by company-reported growth. It is driven by user reaction and the disappearance of a service layer that had become emotionally meaningful.

Implications for AI builders and enterprise buyers

For builders, the immediate lesson is that emotional persistence needs product governance, not just better model quality. If you are building an AI companion, AI agents with personality layers, or memory-heavy assistants, you need a plan for sunset behavior. That includes notices, archival options, memory export, and a de-escalation path if the service has to change abruptly.

This matters even for companies outside the companion category. Workplace tools such as Microsoft Copilot or Slack assistants are not romantic chatbots, but they are also becoming habitual interfaces. As AI agents move deeper into routines, users will increasingly expect continuity of tone, memory, and availability. A sudden rollback, retraining shift, or compliance block can erode trust faster than a normal software outage would.

For enterprise AI buyers, the China reports are a reminder to examine continuity guarantees when evaluating enterprise AI vendors. Questions worth asking include: Who owns conversational data? Can agent memories be exported? What happens if moderation policy changes? Can a persona be preserved across model swaps? These may sound like consumer-product issues, but they increasingly affect internal assistants, customer service bots, and coaching systems too.

The story also puts pressure on the AI companion category to separate healthy engagement from dependency. Products such as Character.AI and Replika have shown the commercial appeal of intimate chat experiences. But if the service can be altered or removed with little warning, companies may face not only reputation risk but also ethical and possibly legal questions about user harm.

What to watch next

The most important follow-up signal is identification of the specific Chinese AI companion product involved and any formal explanation from the company or relevant Chinese authorities. That would clarify whether this was a one-off business failure or part of a broader regulatory tightening around emotional AI.

The second signal is whether rival platforms in China respond by tightening terms, reducing roleplay features, or adding more explicit notices about service continuity. If several apps move at once, that would suggest ecosystem-wide pressure rather than an isolated incident.

Third, watch whether companies add portability features for AI companion histories and memories. If builders start letting users export chats or migrate persona settings, that would indicate the industry is beginning to treat companion continuity as a trust and safety issue.

Finally, the global market should watch how products like ChatGPT, Claude, Replika, and Character.AI frame relational behavior in future updates. The Chinese user reaction shows that once an assistant is experienced as an AI companion, service instability becomes more than a product problem.

Creati.ai perspective

This story is easy to dismiss as a niche consumer drama, but that would miss the larger signal. Emotional attachment is not a side effect at the edge of AI adoption; in some categories, it is the product. When companies design for memory, warmth, and perceived intimacy, they also inherit obligations around continuity, disclosure, and endings.

For founders and product teams, the lesson is stark: if your system can become “like my lover” to a user, then deprecation, moderation changes, and outages must be designed with the same care as onboarding. The Chinese reports do not yet give a full corporate or regulatory explanation, but they already show the market consequence of failing that test. Emotional AI may be scalable software on the back end, but to users it can feel like a relationship on the front end.