Generative artificial intelligence is rapidly penetrating consumer scenarios, and shopping behavior is advancing from “search and recommendation” to an era of “agentic transactions” in which AI assistants compare prices, place orders and even make payments on behalf of users. With AI now at the front end of transactions, the question of who bears responsibility — the platform, the brand or the consumer — when the wrong product is bought or an agent exceeds its authorized scope has become the most pressing operational risk in the eyes of retailers and e-commerce operators.
Wu Xinyang, the newly appointed chairman of the Retailers Association of the Republic of China (Non-Store) and vice chairman of Shin Kong Mitsukoshi, on November 9 chaired his first “Industry Development Forum” since taking office, leading representatives from more than 30 board and supervisory companies spanning the retail ecosystem in face-to-face discussions with Taiwan’s Minister of Digital Affairs Lin Yijing and his Administration for Digital Industries team. The two sides agreed to establish a regular communication mechanism and settled on future directions for cooperation on issues including liability for AI agent shopping, fair competition in cross-border e-commerce, fostering a retail AI ecosystem and joint cybersecurity defense.
Speaking at the forum, Wu said the proposals raised by participants were not the needs of any single company but the shared voice of the entire industry chain. He summarized three major shifts rewriting the rules of retail competition: first, consumers are moving from searching and comparing prices themselves to delegating transactions to AI agents; second, overseas platforms are entering the market with enormous resources, leaving Taiwanese operators that strictly comply with the law at a disadvantage in an unequal competitive position; and third, fraud and brand impersonation tactics are constantly being reinvented, making it difficult for companies to cope alone.
AI Agent Liability: Association Urges Government to Get Ahead of the Curve
The most closely watched focus of the forum was liability for AI agent shopping. Gu Yuanhong, vice chairman of the association and general manager of momo.com (8454.TW), suggested the government plan ahead and prioritize clarifying three sets of rules — authorization scope, liability allocation and AI agent identity verification — to avoid patching holes only after disputes arise.
Lin responded positively to the request, saying Taiwan’s Ministry of Digital Affairs will formally study a “risk framework for AI applications in the e-commerce industry” and will broadly incorporate the association’s practical input as the basis for policy formulation.
Competitive conditions for cross-border e-commerce were another front. Li Guangyao, general manager of the commercial business group at Microsoft Taiwan, called on the government to establish minimum compliance standards for cross-border e-commerce so that domestic and overseas operators are subject to the same set of rules. Lin stated explicitly that “domestic and overseas operators will be treated alike,” stressing that cross-border e-commerce operators “must pay the taxes they owe, undergo the inspections they are required to undergo, and carry the insurance they are required to carry.”
He also disclosed that Taiwan’s Ministry of Digital Affairs has joined hands with the Ministry of Health and Welfare, the Ministry of Economic Affairs and the Ministry of Finance to review exemption rules for product inspections. In the future, product inspections will be strengthened, honest declaration through EZ Way will be enforced, customs clearance will be accelerated for operators with good records, and inspection frequency will be increased for those with poor records.
On age verification, Gu proposed using technology tools to verify age and allowing sales channels with verification capabilities to pilot online alcohol sales. Lin said Taiwan’s Ministry of Digital Affairs’ digital credential technology can verify only that a person is “at least 18 years old” without disclosing additional personal information, making it technically entirely feasible, and that inter-ministerial consultations will continue to build social consensus.
Joint Cybersecurity Defense Built on RISAC Intelligence Sharing
On cybersecurity and anti-fraud, Wu said the association will promote joint defense based on the RISAC retail cybersecurity intelligence-sharing mechanism, linking the early-warning capabilities of industry peers. Lin warned that cases have recently emerged of hackers using agentic AI tools to automatically scan websites for vulnerabilities, urging operators to patch immediately upon receiving alerts. He stressed that after companies proactively report cybersecurity incidents, the role of Taiwan’s Ministry of Digital Affairs is to “prioritize helping companies solve problems rather than issuing fines.”
Addressing the resource bottlenecks companies commonly face in adopting AI, Lin explained that the government offers five major tools — computing power, data, talent, marketing and funding. Retailers can apply for free computing power, and an AI matchmaking event dedicated to e-commerce and retail will be organized, while talks are under way with Taiwan’s National Development Council on a fund worth NT$10 billion (about US$303 million).
Wu concluded by summing up the forum’s consensus in eight words — “fairness, implementation, joint defense and sustainability” — and reiterated that the association is willing to serve as a bridge between the government and industry on policy opening, regulatory relaxation, cybersecurity and anti-fraud intelligence sharing, and talent cultivation, working with Taiwan’s Ministry of Digital Affairs to make policies deliver real results.
Threefold Pressure on Retail Seen Through Four Proposals
The forum highlighted that Taiwan’s retail industry is simultaneously under threefold pressure from technology, competition and public safety. On technology, AI agents are partially transferring transaction decision-making power to algorithms, yet the current consumer dispute resolution framework has no rules designed for “non-human ordering.” On competition, overseas platforms are pressing in with resource advantages while Taiwanese operators are constrained by stricter compliance costs. On public safety, brand impersonation and fraud tactics continue to evolve, and the defensive depth of any single company is clearly insufficient.
The four proposals put forward by the association — covering industrial policy and regulatory relaxation, AI innovation and implementation transformation, cybersecurity protection and anti-fraud joint defense, and retail ESG and digital talent — effectively placed the pain points of the retail industry on the negotiating table all at once. Judging from the response of Taiwan’s Ministry of Digital Affairs, the AI application risk framework, cross-border e-commerce compliance standards and a friendlier cybersecurity reporting mechanism are the three projects most likely to be prioritized in the near term.
For investors, if AI agent shopping becomes institutionalized, demand for identity verification, authorization management and transaction record storage will rise in tandem, potentially benefiting operators with cybersecurity and identity verification technology capabilities. Conversely, before liability rules are clarified, the potential dispute costs platforms bear for AI agent transactions remain a variable that must be factored into valuations of the e-commerce sector.