At WebX 2026 in Tokyo on July 13, former Taiwan Digital Minister Audrey Tang and legislator Ju-Chun Ko revealed the philosophy behind Taiwan's new Virtual Asset Service Provider (VASP) Act and AI Basic Law, passed in the first half of 2026. Tang argued the shift is from 'ambiguity to clarity,' not from conservatism to liberalization, and that delayed regulation always benefits incumbents.
From Gray Zones to Clear Rules
Tang noted that while Bitcoin could be bought at convenience stores in Taiwan as early as 2016, regulatory progress stalled for a decade. The VASP Act, passed in June 2026, and the AI Basic Law, effective January 2026, now change that. 'A gray zone quietly empowers those who can already move within it,' Tang said, emphasizing that the new laws remove the 'interpretation monopoly' that a few players held.
Latecomer Advantage and Regulatory Design
Despite leading global semiconductor manufacturing, Taiwan lagged behind the EU in AI regulation. Ko asked how Taiwan could avoid stifling innovation. Tang drew from the Greek root of 'cybernetics' (kybernetes, or helmsman), arguing that steering speed matters more than acceleration. 'Good regulation isn't about perfect AI settings today,' she said, but about adaptable risk profiles. She praised Japan's community-specific integration types, where AI enters human loops rather than the reverse.
Civic AI: Data as Soil, Not Oil
Tang introduced her 'Civic AI' framework (civic.ai), which treats data as soil rather than oil. Instead of extracting data from communities into giant cloud models, each community maintains its own 'commi'—a small knowledge management system combined with AI. Tang revealed her commi, JTEAMI, runs on Ethereum as agent ID 22714 under ERC-8004, a permissionless agent protocol. It passes four tests: portability across providers, inspectability of routes including smart contracts, community-governed deployment, and accountability in failures.
Proof of Personhood and Proof of Agent
Tang addressed deepfake risks by noting that human identity verification is largely solved through local biometrics and signatures. The next challenge, she said, is 'proof of agent'—proving that an AI agent is genuinely authorized by its human principal. Formal verification tools like Lean and Dafny can cryptographically prove invariants in AI agent code, similar to smart contract hardening. 'Frontier models come and go, but data as soil and steering capacity accumulate,' Tang said.
Bipartisan Tech Policy
Ko, an opposition lawmaker, stressed that technology policy must be transparent and bipartisan. 'I sit in the opposition, you serve as cyber ambassador for the current administration, yet together we advanced Taiwan's AI Basic Law and VASP Act,' he said. Taiwan's journey from regulatory laggard to a clear legal framework for both AI and crypto offers lessons for Japan and other Asian nations.