VIEWPOINT: Why AI is only as trustworthy as the system around it

Dr Tan Dunlin Artificial intelligence is becoming part of aviation’s operational backbone. Across the region, AI is already helping optimise aircraft sequencing, improve predictive maintenance, enhance situational awareness and support more complex air traffic management operations. As AI becomes more deeply embedded in these environments, the real question is no longer how capable AI has become. It is whether AI can be trusted when safety, security and operational continuity are on the line. That distinction matters because aviation evaluates technology differently from most industries. Consumer AI applications can recover from occasional mistakes. Mission-critical systems cannot afford unpredictable behaviour. When AI supports pilots, controllers and airport operators, trust is not a useful extra. It is an operational requirement. The standard changes once AI enters critical systems Much of today’s AI conversation focuses on the model. Larger foundation models, more sophisticated reasoning and higher benchmark scores continue to redefine what AI can achieve. These advances matter, but they are not how aviation evaluates new technology. Airlines, airports and air navigation service providers are less concerned with whether an AI model performs well under ideal conditions than whether it continues operating safely when weather deteriorates, communications are disrupted, sensor inputs change unexpectedly or decisions must be made within seconds. Deploying AI in aviation requires more than technical performance. It requires systems that behave predictably, remain resilient under changing conditions and support human operators when decisions matter most. Trust is built into the system, not just the model AI never operates in isolation. In aviation, it works alongside radar and sensor data, avionics, communications networks, cybersecurity systems, air traffic management platforms and, ultimately, the people responsible for operational decisions. The performance of the overall system depends on how these elements work together under real operating conditions. The real question is therefore not whether an AI model can be trusted. It is whether the operational system remains trustworthy once AI becomes part of it. Building trusted AI goes beyond improving model accuracy. It requires trusted data, resilient communications, secure system architecture, rigorous validation and meaningful human oversight working together as a single operational capability. This is fundamentally a systems engineering challenge. That is also why AI for critical systems cannot be treated as a standalone software layer. It must be engineered into the wider environment in which it is used, so that it remains explainable, traceable and resilient when conditions change. Singapore is helping define trusted AI for operational environments Earlier this year, the government identified aviation as one of Singapore’s National AI Missions, recognising that challenges such as aircraft sequencing, air traffic management and airport operations can become proving grounds for AI solutions with global relevance. As Changi Airport expands with Terminal 5 and passenger traffic continues to grow, the focus is not simply on increasing capacity, but on ensuring increasingly automated operations remain safe, resilient and efficient. Singapore’s integrated aviation ecosystem also provides a unique environment where regulators, airlines, airport operators, air navigation service providers and technology companies can develop and test new operational concepts together. Within this ecosystem, cortAIx SG, Thales’ AI accelerator for mission-critical environments in Asia, is helping accelerate the research and industrialisation of trusted AI for mission-critical environments. Supported by the Singapore Economic Development Board (EDB) and aligned with Singapore’s National AI Strategy, it brings together AI expertise with operational knowledge across aviation, defence and public safety. The value of this approach lies in bringing technology developers together with regulators and operational users early in the development process, allowing new capabilities to be validated in the environments where they will ultimately be used. Trust will define Southeast Asia’s next phase of AI adoption The challenges Singapore is addressing are increasingly relevant across Southeast Asia. As countries invest in airport expansion, modernise air traffic management and strengthen regional connectivity, AI will become an integral part of managing more complex aviation operations. The opportunity extends beyond adopting AI to improve efficiency. It is about deploying AI in ways that strengthen the safety, resilience and reliability of increasingly interconnected aviation systems. That requires close collaboration between technology providers, regulators and operational users, ensuring new capabilities are tested and validated before they become part of day-to-day operations. Singapore’s experience shows that trusted AI is not built through technology alone. It is developed through collaboration across the aviation ecosystem, creating solutions that can be adapted and scaled as the region’s aviation networks continue to grow. The next chapter of AI AI will become an increasingly important part of aviation as airports become more connected, air traffic grows and operational environments become more complex. Across Southeast Asia, the opportunity extends beyond adopting AI to improve efficiency. It is about deploying AI in ways that strengthen the safety, resilience and reliability of the systems people depend on every day. For aviation, trust is not a feature that can be added after deployment. It must be engineered into the entire system through trusted data, resilient communications, rigorous validation, cybersecurity and meaningful human oversight. Only then can AI support operational decisions with the confidence that critical systems demand. (NOTE: Dr Tan Dunlin is Director of Thales Research & Technology and Head of cortAIx Singapore.) For Editorial Inquiries Contact: Editor Matt Driskill at [email protected] For Advertising Inquiries Contact: Head of Sales Sally Passey at [email protected]