Big-O at the 2026 Presidential Hackathon: proposing "Food Safety AI" so anyone can ask one question and know whether today's food is safe
The 2026 Presidential Hackathon theme was "Smart Agriculture, Safety and Sustainability". Big-O Information Intelligence Corp. submitted "Food Safety AI", turning Taiwan's open government data on inspections, traceability and recalls into answers citizens can understand and AI can cite with verifiable sources. Its backend design, AgriTrust Engine, is a trusted-data layer for Taiwan's agricultural AI.
— Big-O Information Intelligence Corp.
Taiwan's Presidential Hackathon has run since 2018 under the guidance of the Office of the President and organised by the Executive Yuan. The 2026 edition, themed "Smart Agriculture, Safety and Sustainability", accepted proposals from 31 July to 31 August. Big-O Information Intelligence Corp. formed the team "Food Safety Guardians" and submitted "Food Safety AI" (proposal no. 215). It received 103 votes in the public quadratic vote (1–10 September) and was not among the 20 shortlisted teams announced on 17 September; the team received the hackathon's commemorative gift, a bamboo-fibre cup. This article records what we proposed, why, and how the ideas flow back into our products.
1. The problem is not "does the government have data" but "can people use it right now"
When a food-safety incident breaks, the first question is "can I still eat what I bought yesterday?". Taiwan's agencies already publish inspection results, production traceability, product recalls, prices, carbon footprints and weather data, but across many sites and datasets. Ordinary people do not start from dataset names, and messages forwarded in LINE groups or answers from generative AI may rest on old news or foreign sources.
"Food Safety AI" is a public Q&A and fact-checking service built on open government data. A user asks in plain language — "are strawberries safe lately?", "what about strawberries from Miaoli?", "a LINE message says a vegetable has ten times the pesticide limit, is that true?" — and the system assembles recent official records, shows source, update time and confidence, and says plainly when the data are insufficient. A "Daily Food Safety Radar" on the home page turns recent official data into green, yellow, red and grey states that elders and family carers can read at a glance.
2. The engine behind it: five modules of AgriTrust Engine
The front end speaks plainly; the back end speaks trusted-data governance. AgriTrust Engine turns scattered government data into an evidence chain that AI can understand, cite and verify:
1. Evidence Collector — fetches agricultural, food-safety, weather and price data on a schedule and records source and version.
2. Trust Graph — links items, origins, incidents, inspections, labels, time and sources so the AI knows which record answers which question.
3. Evidence Passport — keeps the source URL, data version, retrieval time, hash and integrity information for every published fact, so every answer can be traced to where the data came from and when it was updated.
4. AI Cross-check — uses multiple models and rules to detect insufficient sources, stale data or high-risk inference.
5. Feedback Loop — observes gaps in AI answers and feeds them back into data structure improvements before re-testing.
Timestamping (for example OpenTimestamps) only provides integrity and time proof in this design; it never claims that blockchain alone proves data came from the government. Source trust rests on official APIs, official URLs and data versions.
3. Why not just another chatbot
The innovation is not answering food-safety questions with AI; it is the closed loop of public data → AI answer → citizen query → error detection → data improvement → re-verification. Instead of a query portal or a chat skin over search, the proposal measures whether AI understands and cites correctly, says "insufficient" when it is, and pulls every judgement back to official evidence. Phase one uses only lawfully open, reusable data and touches no sensitive personal data.
4. Outcome and what we learned
The proposal received 103 votes in the public quadratic vote and was not among the 20 shortlisted teams. The team received the hackathon's commemorative gift: a bamboo-fibre cup made from 100% natural material, free of plastic and PLA — fitting for a theme of safety and sustainability.
Not being shortlisted does not diminish the idea. Three of its building blocks are already in our products: the Evidence Passport (source, version, time, hash) maps to the anchored, publicly verifiable records of DSCS; AI Cross-check maps to how EcoGEO Audit tests brand citations across four generative AI systems every month; and "say insufficient when it is" is the common rule for all our AI features. We will keep following updates to Taiwan's open food-safety data and revisit the topic when a suitable field partner appears.
5. About Big-O
Big-O Information Intelligence Corp. was founded in November 2023 in Neihu, Taipei. It builds the Big-O AI Brand Infrastructure Platform and serves SMEs and brands with a SaaS + AaaS model; its six products are EcoGen AI, EcoComm AI, EcoGEO Audit, EcoCare AI, AI Co-Worker and DSCS, and its core technology holds Taiwan utility patent M683041. Big-O exhibited at the Taipei Digital Application Expo in 2025 and 2026, the Greater South AI Smart Health Exhibition and TAOYUAN FUTURE PEAK x BIOMED-X, was featured by DIGITIMES, is a Taipei City Youth Development Office grantee, and is one of nine Taiwan companies taking part in MODA's Green Software Label verification. The hackathon proposal is a public-service concept not yet implemented; existing products served only as evidence of the team's technical capability.