Introduction
Our team advises a broad range of technology companies on regulatory compliance in Taiwan. One area of increasing client concern involves companion chatbots—AI tools that simulate emotional intimacy—and the risks they pose to minors whose minds are not yet fully mature. Recent studies reveal that these chatbots can be easily manipulated to generate highly inappropriate content, including sexual innuendo, violence, and the encouragement of self-harm. Tragically, reports have also linked minors’ emotional dependence on AI companion chatbots to psychological distress and, in extreme cases, suicide.
Faced with these growing concerns, regulators around the world are working to establish guardrails. In this article, we review recent legislative efforts in California and compare them with Taiwan’s Artificial Intelligence Basic Act (the “AI Basic Act”), exploring what these developments may mean for companies operating in or expanding to Taiwan.
Finding the Balance: Risk Management vs. Prohibition
California lawmakers recently debated two distinct approaches to regulating companion chatbots: strict risk management versus an outright ban.
On October 13, 2025, California Governor Gavin Newsom signed Senate Bill 243 (“S.B. 243”) into law, establishing a “risk management” model. The law requires operators, among other things, to clearly disclose a chatbot’s identity as artificial rather than human, establish mechanisms to prevent references to suicide or self-harm, and implement special protective measures for minors. Crucially, S.B. 243 mandates that starting July 2027, companies must submit an annual “crisis intervention report” detailing their efforts in detecting suicidal ideation and the preventative measures deployed.
Conversely, Assembly Bill 1064 (“A.B. 1064”) attempted to create a presumptive ban on companion chatbots for minors unless operators could prove that their technology was harmless. The governor ultimately vetoed this bill, noting that its broad language could inadvertently deprive minors of the educational and social benefits offered by AI tools. This veto sends a strong signal to the tech industry that regulators currently prefer implementing safety guardrails rather than imposing outright bans.
Taiwan’s AI Basic Act
Following these global trends, Taiwan passed the AI Basic Act on December 23, 2025, which was officially promulgated and came into force on January 14, 2026. Rather than dictating rigid, itemized rules for specific technologies, the AI Basic Act serves as a high-level foundation for Taiwan’s AI industry.
Much like California’s S.B. 243, Taiwan’s framework provides that the government shall prevent AI applications from creating illegal situations and prioritizes the interests of minors by requiring enhanced transparency and warning labels for high-risk applications. Furthermore, it grants regulators the authority to establish a comprehensive risk classification system. Notably, Taiwan’s AI Basic Act provides that if an AI application’s risks cannot be effectively managed, authorities must restrict or prohibit it—an approach that mirrors California’s vetoed A.B. 1064. To ensure accountability, the framework also requires clear liability, remedy, and insurance mechanisms for high-risk AI applications.
Conclusion
While Taiwan’s AI Basic Act operates as a high-level guideline for now, the regulatory trajectory is clear. Driven by its risk classification mandate and its focus on protecting minors, regulators will likely classify companion chatbots as “high-risk” in forthcoming supplemental legislation. Tech companies looking to expand in Taiwan should prepare for strict new rules regarding the protection of minors, suicide and self-harm risk detection, and mandatory transparency. Although California’s A.B. 1064 was ultimately vetoed, its underlying focus on safeguarding minors remains a valuable reference point for Taiwan as it seeks to balance innovation with the physical and mental wellbeing of its citizens. We recommend that companies closely monitor developments in subsidiary legislation and begin planning their compliance strategies accordingly.
This is an abridged translation of the original article in Mandarin written by Gary Kuo, Chi-hsien Nieh, and Jung-chen Lee, which can be found here. Because the content of this article involves foreign regulations that are subject to rapid change, we recommend that readers independently consult original sources when citing the information herein.
If you have any questions or require additional information on AI regulation and compliance in Taiwan, please feel free to contact us at gkuo@winklerpartners.com and cnieh@winklerpartners.com.