WebX 2026 debuted in Tokyo on July 13, featuring a special dialogue between Taiwan lawmakers Kuo Ju-yun (葛如鈞) and former Taiwan Minister of Digital Development Audrey Tang (唐鳳). The discussion focused on two major regulatory framework projects completed by Taiwan in the first half of 2026: the Virtual Asset Service Provider Act (VASP) and the AI Basic Act. Audrey Tang discussed the next phase direction for Taiwan’s digital governance, covering regulatory philosophy, the “late-mover advantage,” civic AI, AI agent identity, and blockchain identity.
VASP Act and the AI Basic Act: Two major digital governance frameworks Taiwan completed in the first half of 2026
According to reports, Taiwan completed two legal systems in succession in the first half of 2026 as follows:
VASP Act (Virtual Asset Service Provider Act): Passed in the third reading on June 13, 2026. It establishes a registration system for virtual asset businesses and incorporates requirements for customer protection, anti–money laundering, and operational management.
AI Basic Act: Took effect on January 14, 2026. With 20 principle-based articles, it clearly lists seven major principles: sustainability, human autonomy, privacy, cybersecurity, transparency, fairness, and accountability.
Audrey Tang pointed out, “Society has not yet formed consensus” may sound neutral, but in practice it often gives an advantage to the side that already has the capability to act. Regulatory delays allow vested interests to continue to benefit. The key focus of this round of institutional shift is to change the interpretive space originally controlled by a few people into public rules that are transparent, discussable, and enforceable.
Audrey Tang’s “late-mover advantage” and philosophy behind the regulatory shift
Based on Audrey Tang’s remarks at WebX 2026, the advantage of late movers is that they can learn from other countries’ regulatory experiences while avoiding locking fast-changing technologies into rigid classifications. She explained the governance core using the Greek etymology behind the term “cybernetics,” saying governance is not about pursuing acceleration but about “retaining the ability to change direction.”
Audrey Tang argues that good AI regulation should not only ask whether today’s system classification is perfect. Large models have entered a new stage where they can be used to train smaller models; fixed tiering will quickly become inaccurate. Regulation should adjust based on risk patterns and allow each community to adopt AI in line with its own needs.
She also mentioned Japan’s approach of creating different integration types based on community differences, which provides an important reference for policy design in Asia.
civic.ai, AI agent identity, and on-chain approval
According to reports, Audrey Tang introduced the “civic.ai” architecture during the dialogue and put forward the view “Data is soil and never oil.” She argues that data should be treated as soil that nourishes community knowledge, not as a resource that can be extracted from communities and delivered to massive cloud AI models. Each community can own a “commi” that combines small-scale knowledge management with AI, allowing AI to become a bridge for cross-party and cross-cultural communication.
Regarding AI agent identity, Audrey Tang said: if a certain AI agent represents an individual to carry out trades or make decisions, the system must be able to prove that the agent was indeed authorized by that person. Her commi, “JTEAMI,” is already operating on Ethereum with ERC-8004 and Agent ID 22714. She also emphasized that if an AI agent is to conduct trades on someone’s behalf, it must pass tests such as portability, path verifiability, community-deployed governance, and fault accountability.
FAQ
What are the core requirements of Taiwan’s VASP Act, and when does it take effect?
According to reports, Taiwan’s VASP Act (Virtual Asset Service Provider Act) was passed in the third reading on June 13, 2026. It establishes a registration system for virtual asset businesses and incorporates requirements for customer protection, anti–money laundering, and operational management. The law does not follow a “heavier penalties first” approach; instead, it attempts to set regulatory boundaries between fostering innovation and protecting rights. The specific effective date and implementation details are subject to official announcements by Taiwan’s financial regulatory authorities.
What exactly does Audrey Tang’s “late-mover advantage” argument refer to?
According to reports, Audrey Tang said at WebX 2026 that late movers can refer to other countries’ regulatory experiences and avoid locking fast-changing technologies into rigid classifications, thereby retaining the ability to shift regulation. She believes good AI regulation should dynamically adjust based on risk patterns rather than striving for a fixed “perfect” classification system, because the pace of technological development makes any static tiering easily outdated.
What is ERC-8004, and how is it related to AI agent identity verification?
According to reports, Audrey Tang said her commi “JTEAMI” is already operating on Ethereum with ERC-8004 and Agent ID 22714. ERC-8004 is a proposal on Ethereum for an AI agent identity standard. Audrey Tang stated that if an AI agent is to represent an individual in making trades or decisions, it must pass tests such as portability, path verifiability, community-deployed governance, and fault accountability to prove that the agent’s authorization is legitimate. The specific technical specifications are subject to Ethereum’s official EIP documents.
