InfrastructureIssue 064 records
Transportation
A governance-heavy day with no independent outcome data. Incheon's port authority issued AI rules requiring human oversight, a 30-second manual-override rule for high-risk equipment and fairness audits. Adelaide's AI-assisted parking fines were withdrawn in one of four challenged cases, raising questions about human review. Airport leaders from Hong Kong and DFW stressed cost screening and frontline co-design. CitySwift launched real-time decision support for bus controllers, per the vendor. No U.S. state DOT, tolling or fleet evidence was found.
- Evidence records
- 4
- Cross-source patterns
- 1
- Evidence classes
- 3 independent reporting1 vendor claim
- Outcomes
- 3 emerging1 cautionary
- Source freshness
- 4 new this fortnight
- Research completed
- 2026-11-04
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Human oversight is being specified, and tested, as the key AI control
Incheon mandates human verification and timed manual override, while Adelaide's withdrawal data shows two-person review can still let contestable notices through; oversight needs metrics, not just a rule.
Operating questionWhat metric will show the agency's human review of AI outputs is actually catching errors?
Supporting evidenceIncheon Port Authority (IPA)City of Adelaide
Evidence ledger
Full record · every source keeps its link and limitations
Incheon Port Authority issues AI guidelines with human oversight, manual override and risk tiers
IPA distributed AI guidelines with four principles and eight provisions. Human oversight is mandatory even for autonomous control; high-risk equipment-control AI needs two or more supervisors and, as a rule, a switch to manual control within 30 seconds of malfunction. Uses are tiered: statutory high-impact (lock gates, automatic berthing/mooring; prior impact assessment), priority-managed (berth allocation and hazmat storage recommendations; official approval, periodic revalidation) and general. Allocation AI faces six-monthly fairness audits with disclosed results. A CAIO and AI ethics committee are planned.
Why it matters, evidence and limitations
- Why it matters
- A concrete port-authority governance model that U.S. port authorities and DOTs could compare with NIST AI RMF-style policies, especially for berth allocation fairness and equipment-control override.
- Evidence and measured results
- Policy document as reported by press; no outcomes, audit results or incident data yet. The guideline text itself was not opened.
- Limitations and uncertainty
- Responds to Korean AI Framework Act and AI/data administration act; transferability to U.S. legal context is limited. Reported, not yet implemented outcomes.
Airport executives at Super Terminal Expo stress cost-effectiveness and frontline involvement in AI adoption (HKIA, DFW)
AAHK's CEO said the authority assesses financial feasibility and whether AI strengthens existing systems, optimizes resources or eases manpower shortages, noting substantial initial expenditure. HKIA is piloting smaller-scale projects and building an AI-literate in-house team through training. DFW's Chris McLaughlin said a recent DFW AI project used workshops with over 50 frontline workers, managers and consultants to define functionality, which he credited with higher adoption.
Why it matters, evidence and limitations
- Why it matters
- Includes a U.S. public airport (DFW) describing co-design with frontline staff, and a major airport authority prioritizing cost-effectiveness, relevant to airport CIOs planning AI pilots.
- Evidence and measured results
- Executive remarks only; no project names, metrics or evaluation. Adoption and productivity claims are self-reported.
- Limitations and uncertainty
- Conference statements; titles differ across sources (The Standard calls McLaughlin COO; the HKIA release lists him as CEO). No measured outcomes.
One in four challenged Adelaide Park Safe AI-assisted parking fines withdrawn; oversight questioned
Reporting (citing The Advertiser) says one in four challenged fines flagged by Adelaide's AI-assisted Park Safe camera vehicles has been withdrawn since June 2024. Two staff review each case. The council says some withdrawals reflect mitigating circumstances and that under 1% of 11,068 notices in Q1 FY2026-27 were withdrawn after review. A University of Sydney academic warned human review can become a rubber stamp. In Western Australia about 5,700 of 81,412 AI-camera seatbelt fines were withdrawn on review.
Why it matters, evidence and limitations
- Why it matters
- Direct evidence on how human-in-the-loop review performs in AI-assisted parking enforcement, a growing U.S. municipal practice (e.g., camera-equipped enforcement vehicles).
- Evidence and measured results
- The 25% figure applies to challenged notices, not all notices; the council's under-1% figure uses a different base and period. No breakdown of errors versus mitigating circumstances.
- Limitations and uncertainty
- Tabloid reporting relaying council and academic statements; underlying Advertiser data not inspected; Australian context.
CitySwift launches real-time predictive decision support for bus controllers
CitySwift announced its first real-time product: predictive decision support that analyzes live network performance, predicts how delays affect each vehicle's future trips, ranks at-risk buses and recommends interventions (spare vehicle, cover the next journey, warn passengers). It reuses models of driver behavior, layover, recovery time and traffic, and links to recurring-pattern analysis.
Why it matters, evidence and limitations
- Why it matters
- Relevant to transit agencies' operations control centers considering AI decision support for service recovery.
- Evidence and measured results
- Vendor announcement; no deployment sites, evaluation or measured results for the real-time product.
- Limitations and uncertainty
- Vendor claim; no customers or outcomes named for this product; earlier archived CitySwift London result concerned planning, not this real-time tool.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Vendor claim
- A supplier-provided assertion that has not been upgraded to independent evidence.