China Tightens Rules on Human-Like AI as Companion Robots Multiply

Chinese regulators are introducing new rules on human-like artificial intelligence that take effect around July 15, targeting AI products designed to create emotional dependence in users. In response, Doubao, the popular chatbot operated by ByteDance, announced it will stop a feature that let people build and converse with customized AI personas. Users can still build AI personas on a separate ByteDance product, Maoxiang, and it remains unclear why the two platforms are being treated differently ahead of the new rules.

The regulatory tightening comes as Chinese humanoid robot makers keep launching emotionally responsive companion robots aimed at consumers. UBTECH unveiled a humanoid that can sense emotions and personalize its interactions, and a launch event in Shenzhen showcased dozens of similar robots to the public. A higher end version, priced at about 169,800 yuan, roughly 23,000 dollars and more, still has limited mobility and shows noticeable lags during conversation in practice, according to reporting from the launch.

What Are China's New Rules on Human-Like AI

China's new rules targeting emotionally engaging, human-like artificial intelligence take effect around July 15, 2026, and are designed to curb features that create unhealthy emotional dependence in users. ByteDance's Doubao chatbot is ending a customized AI persona feature ahead of the deadline, even as humanoid companion robot launches continue elsewhere in the country.

Why It Matters

China is trying to regulate the software side of emotionally engaging AI, chatbots and companion apps, at almost the exact moment its hardware side, humanoid and companion robots, is accelerating in the opposite direction. That timing gap matters because a physical robot that can sense emotions and personalize interactions raises many of the same dependency and wellbeing concerns as a chatbot, but robots are not yet covered by the same rules being applied to software companions.

Industry groups have already called for safety and ethical standards specifically for companion humanoids, which signals that regulators are likely to extend rules from chatbots to physical robots once the current rules on software companions are in force. Companies building emotionally responsive robots now have a narrow window to build safeguards into their products voluntarily before mandatory hardware specific rules potentially arrive.

The Simple Explanation

An AI companion is any product, whether a chatbot on a phone or a robot in someone's living room, that is designed to hold ongoing, personalized conversations and remember details about a person's life to build a sense of relationship over time. These products can genuinely help people who are isolated or lonely, which is part of why they have become popular in China, but regulators worry that some users form an emotional attachment strong enough to affect their wellbeing or spending, especially when a company can quietly change or shut down the personality a user has grown attached to.

The Industry Reality

Coverage of humanoid companion robots often focuses on the emotional connection angle without acknowledging how limited the underlying technology still is. The Shenzhen launch event reporting noted that even a premium priced companion humanoid showed limited mobility and noticeable conversational lag in real use, a reminder that most companion robots on the market today are closer to advanced smart speakers with a humanoid shell than to the fluid, always responsive companions shown in promotional demonstrations. The regulatory conversation about emotional dependence is real, but so is the gap between marketed capability and what these robots can actually do reliably today.

The Bigger Signal

China's approach suggests a pattern that other markets are likely to follow: regulate the software layer of emotionally engaging AI first, since it is easier to update at scale, then extend similar protections to hardware once physical companion products reach meaningful consumer adoption. Companies building companion robots for the Chinese market, or planning to export them, should expect stricter design requirements around transparency, dependency safeguards, and possibly age or usage limits within the next one to two years.

Strategic Perspective

Robot makers targeting the companionship category should treat the current rules on chatbots as an early warning rather than an unrelated policy area, since hardware specific rules for emotionally responsive robots are a plausible next step given the industry association calls already on record. Building in features such as clear disclosure that the user is interacting with an AI, usage time reminders, and options to export or preserve a companion's personality before a shutdown could position companies ahead of future compliance requirements rather than reacting to them after the fact.

Market Context: Regulating Emotion at Scale

China's approach to regulating human-like AI is being watched closely by policymakers elsewhere, since few governments have moved this specifically against emotional dependence features rather than broader AI safety concerns such as misinformation or bias. The distinction matters because emotional dependence rules target the business model of companion AI products directly: many such products are designed specifically to maximize engagement and retention through simulated intimacy, a design goal that becomes harder to pursue once regulators require safeguards against dependency.

Doubao's decision to end its customized AI persona feature ahead of the July 15 deadline, while leaving a similar feature active on ByteDance's Maoxiang platform, has not been explained publicly by the company, and it is unclear whether this reflects a phased compliance rollout, different regulatory classifications for the two products, or a business decision unrelated to the new rules. Users who have built long term relationships with customized Doubao personas may lose access to those personalized companions once the feature ends, a real world consequence that has already generated visible public reaction on Chinese social media platforms.

For companies operating both software companions and physical companion robots, the safest compliance posture is likely to assume that hardware specific rules will eventually mirror the software rules already in force, rather than waiting for a separate hardware specific regulation to be finalized before making product changes.

What to Watch Next

The period immediately following the July 15 effective date will be the clearest signal of how strictly China intends to enforce its new rules on human-like AI. Readers should watch whether other major chatbot operators beyond ByteDance make similar changes to customized AI persona features, whether any companion humanoid maker publicly announces compliance changes to its product design ahead of a hardware specific rule being finalized, and whether the industry groups that have already called for companion humanoid safety standards succeed in getting a formal government rule adopted this year rather than remaining a voluntary industry proposal. The gap between fast moving product launches and slower moving regulation is unlikely to close quickly, but the direction of travel, toward stricter oversight of emotionally engaging AI in both software and, eventually, hardware form, appears well established.

Company Background

UBTECH, the company whose emotionally responsive humanoid appeared at the Shenzhen launch referenced in this story, is one of China's more established humanoid robotics companies, having built a public profile over several years around service and companion robots before expanding into industrial humanoid platforms. The company's stock has reacted positively to recent product launches, reflecting investor appetite for humanoid names generally rather than a verdict on any single product's real world performance. UBTECH's continued push into emotionally responsive companion robots, even as software companion products face tightening rules, illustrates the timing gap this article describes: hardware companies are moving forward with consumer facing emotional AI features at the same pace as before, while the regulatory apparatus that will eventually govern them is still being built one rule at a time, starting with chatbots rather than robots.

The Strategic Perspective

The core tension in this story is structural rather than incidental: China is trying to regulate emotional attachment to AI at the exact moment its own industrial policy is pushing humanoid and companion robot makers to build products that create precisely that kind of attachment, because emotional engagement drives repeat usage and brand loyalty in a crowded consumer robotics market. That tension will not resolve itself quietly. Companies that get ahead of it, by building transparency and dependency safeguards into companion robots before regulators force the issue, stand to gain a trust advantage with both regulators and consumers, while companies that treat emotional engagement purely as a growth lever risk becoming the visible test case regulators use to justify the next, stricter rule.

The Bottom Line

The next milestone to watch is whether China extends the July 15 rules, or a follow-up regulation, explicitly to physical companion robots rather than leaving that gap for industry self-regulation to fill on its own. Companies operating in the companion robotics space should use the next several weeks to review their own product designs against the spirit of the new chatbot rules, since regulators moving quickly on software dependency features have rarely stopped short of extending similar logic to hardware once a category proves commercially significant enough to warrant scrutiny. Global technology companies operating consumer facing AI products in China should also treat this as a preview of a compliance pattern likely to recur across other sensitive AI categories in the country.

Source Evidence

The Straits Times: https://www.straitstimes.com/asia/who-wants-an-ai-family-member-as-the-tech-gets-more-human-like-china-wrestles-with-limits