The Trustworthy AI Hackathon in Taipei will run from Aug. 29 to Aug. 31, with Demo Day on Aug. 31, after organizers built the event around roughly 22 problems gathered from three industry expert meetings held on June 7, according to Taiwan Blockchain Enthusiasts Association chairman Taka Kao.
Speaking at a media tea session for the “2026 Trendy Taipei” technology festival, Kao said the hackathon topics were not drafted behind closed doors by the organizers. Instead, they came out of three closed-door meetings with industry representatives and technical experts that lasted a combined 24 hours. Registration closes on Aug. 5, and Kao said nearly 20 teams had already signed up as of the Aug. 4 session, with final participants to be selected afterward.
Questions around AI agents now focus on identity, authority and liability
Kao framed the issue around a series of unresolved questions that come with the spread of AI agents: Who is the counterparty? Which organization does the AI belong to? Who bears responsibility behind it? Was it actually authorized to act? What is the scope of that authorization? Do its statements carry legal force? Can its commitments be enforced? Can the interaction be audited later? If something goes wrong, can evidence be produced, and who is accountable?
“These things actually do not have answers right now,” Kao said.
From Responsible AI to Trustworthy AI
Kao broke the recent evolution of AI into three stages. First, AI helped people write code. Later, it became a tool for writing. Now it is turning into an agent capable of autonomous actions, including signing legal documents, serving as a digital twin, or collaborating across institutions.
He said industry discussions used to center on Responsible AI. The problem has now shifted to Trustworthy AI. The gap between the two, in his view, lies in concrete questions around credential-backed responsibility, data sovereignty, and the allocation of rights and obligations.
The association’s proposed direction is a credential model. Kao said that would combine digital identity with data governance, plus ownership, authorization, and accountability, and package those rights and responsibilities into credentials so future agents can obtain permission and act on behalf of their owners.
He linked that approach to an international example raised at the event. Earlier this year, Jack Dorsey’s Block launched Buzz, which he described as an effort to give every AI agent an “identity passport.” The distinction, he said, is that the Taipei hackathon is built directly on practical problems faced by local industries.
About 22 problem statements came out of 24 hours of closed-door sessions
Kao stressed that the challenge design was not an exercise in internal brainstorming. The association first studied how different countries approach credential trust frameworks, then held three closed-door expert sessions on June 7 for a combined 24 hours. From those discussions with industry and technical participants, organizers distilled about 22 questions.
He offered several examples at the event:
- Cross-border business credentials: The LEI standard already exists internationally, but Taiwan’s current business credentials cannot yet be used abroad.
- Carbon verification: When companies submit materials tied to carbon border requirements, they may expose trade secrets. The question is whether zero-knowledge proofs can support compliance without revealing confidential business information.
- E-commerce anti-fraud checks: Organizers asked whether sellers can undergo real-name verification without disclosing personal ID numbers, and whether those unable to obtain credentials should be treated as suspect first.
- Auditing agent authorization: When AI agents move across the internet and perform actions, how can others verify that a human owner actually authorized them? Kao used the example of authorizing an agent to buy TSMC shares and asked whether that specific trade could later be proven to have been authorized.
The common thread, he said, is that these are not hypothetical future issues. Companies are already filing carbon disclosures, and they are already dealing with fraudulent links. What is still missing is an auditable mechanism.
Schedule, prize pool and pre-event workshops
The Trustworthy AI Hackathon will be held at N24 Taipei Ark in Nangang, Taipei, from Aug. 29 to Aug. 31, with Demo Day on Aug. 31. Registration closes on Aug. 5.
Under information released earlier by the organizers, total prize money starts at $12,000. One first-place team will receive $5,000, two second-place teams will receive $2,000 each, and three third-place teams will receive $1,000 each. Up to 20 teams will be admitted, and the event is opening individual registration for the first time.
Two workshops are scheduled before the competition. The Aug. 15 online session will cover verifiable credentials, privacy-preserving computation, and agent engineering. The Aug. 22 in-person session will feature problem briefings from the industry sponsors behind the challenge topics, followed by team-by-team clinic sessions.
“The spirit of our association has always been hands-on. We like to do things directly and get started right away,” Kao said.
First of four events under the 2026 Trendy Taipei technology festival
Kao said the hackathon also serves as a lead-in to the 11th Blockchain Enthusiasts Annual Conference, which is scheduled for November.
According to the event briefing, the Trustworthy AI Hackathon is the first of four activities under the “2026 Trendy Taipei” technology festival. The lineup will continue with AFA Awards on Sept. 1, FinTechOn 2026 and the AFA Summit on Sept. 2 and Sept. 3, and Taiwan Future Festival FUTUREMODE from Sept. 4 to Sept. 6.
What problem the hackathon is trying to solve
Based on the organizers’ description at the media session, the hackathon is focused on identity and accountability for AI agents. As agents gain the ability to sign, transact, and collaborate across institutions, the central questions become whether they belong to a specific organization, whether they are properly authorized, whether their commitments have legal effect, and whether their actions can be audited and evidenced afterward. The association’s answer is to use verifiable credentials to handle authorization and accountability.
This report is based on on-site coverage from the “2026 Trendy Taipei” technology festival media tea session cited in the source report.