Aviation safety – Air Marshal's Perspective

October 18, 2025October 21, 2025

761: AI AND MILITARY AIRCRAFT AUTOMATION: BALANCING SAFETY WITH CAPABILITY

Artificial intelligence (AI) and automation are revolutionising military aviation. These technologies enable maximum operational capability through autonomous flight, real-time decision-making, and enhanced resource management. They also raise significant safety concerns, including system reliability, ethical considerations, and the need for continuous human-AI interaction. Achieving an optimal balance between enhancing capability and ensuring operational safety is essential. This requires rigorous testing, adaptive standards, and human oversight to ensure mission success and promote safety.

Capabilities Enhanced by AI and Automation

Automation is transforming military aviation by adding new capabilities, enhancing combat effectiveness and efficiency.

Autonomous Operations and Swarm Tactics. AI enables autonomous take-off, navigation, and landing even in hostile or GPS-denied environments. Projects such as the U.S. Department of Defence’s Replicator vision of sending thousands of autonomous vehicles, including drones, on deployment by 2026. They intend to employ swarm intelligence to be utilised for reconnaissance, targeting, and swarming enemy defences. Boeing’s MQ-28 Ghost Bat is an example of a system that augments manned fighters by carrying out reconnaissance and engaging threats independently, de-loading pilot workload. India’s Combat Air Teaming Systems (CATS) and Rustom UAVs use sensor fusion technology, so that manned and unmanned platforms can work together in real time to attack and defend against threats.

Predictive Maintenance and Logistics. Predictive maintenance with AI analyses data from aircraft engines to predict failures, maintaining optimal scheduling and fleet availability. Digital twins, or virtual replicas that account for wear, damage, and flight history, allow faults to be preemptively identified before they occur. A 30% reduction in downtime and millions of dollars in savings can be achieved. The Air Forces and others have utilised these systems to improve logistics and strategic readiness, with aircraft still mission-effective.

Navigation and Decision Support. AI routes for safety and fuel optimisation. AI in emerging fighters such as DARPA’s Air Combat Evolution (ACE) program assists pilots with real-time battlefield analysis and threat identification. This aids faster and more accurate decisions. For instance, AI-controlled F-16s have executed high-speed manoeuvres exceeding 550 mph, responding to dynamic combat scenarios in increments of a fraction of a second.

Command and Control Improvements. The US Joint All-Domain Command and Control (JADC2) employs AI to enable unfettered sharing of information across air, land, sea, and cyber domains. This enables man-machine collaboration for rapid and precise decision-making. AI systems such as the XQ-58A Valkyrie demonstrate autonomous reconnaissance, jamming, and strike operations. They are force multipliers in network-centric warfare. These innovations disrupt the power balance, enabling a rapid response against emerging threats.

Safety Risks and Challenges

Just as AI enhances competence, it poses real threats that must be dealt with in order to promote safe functioning.

System Reliability and Failures. AI’s adapting behaviour can result in unpredictable effects, i.e., errors or bias, during exceptional incidents. Past software failures in military systems have led to accidents, and poor testing increases the potential for these effects. Premature deployment of unmanned systems can result in unforeseen lethal outcomes, i.e., in actual drone crashes during the Ukraine wars.

Ethical and Stability Implications. Autonomous systems can misinterpret circumstances, possibly worsening conflict or jeopardising global stability. Moral dilemmas arise with AI-generated lethal decisions, notably responsibility dilemmas under international humanitarian law. The swift proliferation of autonomous drones addresses actual threats in the world and not alleged dangers such as bioterrorism.

Certification and Regulatory Gaps. Current standards, such as DO-178C and MIL-HDBK-516C, do not fully account for AI’s adaptability. This creates challenges in validation and exposes hardware vulnerabilities. Unlike civil aviation, military applications often experience inconsistent safety compliance, complicating certification for AI-driven systems.

Human Factors. There can be an overdependence on AI, causing pilot proficiency to be lost, particularly in manual flying and quick decision-making. Control handover between human pilots and AI may be challenging in a crisis. There can be automation bias that causes pilots to ignore critical cues. New ideas, e.g., AI-checked conditions of ejection seats and well-being of the pilot, are thrilling but require scrupulous application so that it does not create unforeseen problems.

Cybersecurity Threats. Military aircraft powered by AI are vulnerable to hacking, spoofing, and adversarial attack. These can invalidate important systems and bring about disastrous failures. Cybersecurity plays an important role in maintaining operational integrity.

Balancing Capability with Safety: Strategies and Frameworks

Various measures are being taken by military forces across the globe to contain risks and maximise benefits from AI.

Strict Testing and Phased Introduction. Projects such as Replicator and DARPA’s ACE target strict testing in complete simulations to predict infrequent events and provide reliability prior to deployment. Phased integration within simulated areas provides additional robustness. Autonomy training conducted by the U.S. Air Force employs onboard sensors for enemy detection, while periodic manual flight and emergency procedure training maintain pilot proficiency.

Human-in-the-Loop Systems. Human control over major decisions, particularly the application of force, is important for secure integration of AI. AI is used as a co-pilot and never a replacement, with override rights still under human pilots. For example, autonomous jet test flights like those for the XQ-58A Valkyrie include standby pilots to ensure control.

Redundancy and Fail-Safes. Various safety features, such as manual reversion modes and fallback emergency provisions, enable pilots to regain control when AI systems fail. Tough validation procedures, as those in place for Helsing’s Centaur agent and its interaction with Saab’s Gripen E, enable AI to integrate with installed systems securely.

Certification Standard Development. The development of a systematic safety approach to AI-critical systems involves reviewing existing standards, such as MIL-HDBK-516C and the EASA AI Roadmap, conducting a gap analysis to identify where weaknesses lie, iteratively revising standards to incorporate AI-specific conditions, and examining them in depth to remove overlaps and new requirements. It adapts civil and military systems to deliver effective verification, validation, and continued airworthiness for AI systems.

Talent Development and Recruitment. Artificial intelligence technologies for weather forecasting, maintenance, and operational decision-making enhance readiness through optimising training. Hire AI specialists to monitor and refresh high-risk models under strict testing to provide long-term reliability and safety.

Conclusion

Military aviation is being transformed by artificial intelligence and automation. They provide capabilities that have never been seen before in terms of autonomy, decision-making, and logistics. They bring significant safety, ethical, and strategic problems, too. The future relies on man-machine collaboration, where AI augments human decision-making and not substitutes it. Through constant testing, adaptive certification standards, robust cybersecurity, and ethical governance, militaries are able to leverage AI potential while reducing risks. Ongoing global forums, such as 2025 panels, present cooperation and human control across the globe to ensure AI assists airpower responsibly, balancing capability and safety in driving sustainable advancement.

Please Add Value to the write-up with your views on the subject.

For regular updates, please register your email here:-

Subscribe

References and credits

To all the online sites and channels.

Pics Courtesy: Internet

Disclaimer:

Information and data included in the blog are for educational & non-commercial purposes only and have been carefully adapted, excerpted, or edited from reliable and accurate sources. All copyrighted material belongs to respective owners and is provided only for wider dissemination.

References:-

Cummings, M. L. (2017). Artificial intelligence and the future of warfare. Chatham House.

Eraslan, E., Yildiz, Y., & Annaswamy, A. M. (2019). Shared control between pilots and autopilots: Illustration of a cyber-physical human system. IEEE Transactions on Human-Machine Systems, 49(5), 436–447.

Heydarian Pashakhanlou, A. (2019). AI, autonomy, and airpower: The end of pilots? European Security, 28(4), 523–538.

Hobbs, K. L., & Li, B. (2023). Safety, trust, and ethics considerations for human-AI teaming in aerospace control. Journal of Aerospace Information Systems, 20(6), 280–293.

Jurado, R. D. A. (2024). The current state of standardisation of AI for civil and military aviation. Safety Science, 169, 105178.

Jurado, R. D. A. (2025). Enhancing safety in military aviation: A systematic approach to the development of AI certification standards. Aerospace, 12(1), 72.

Kirwan, B. (2024). The impact of artificial intelligence on aviation safety culture. Aerospace, 11(10), 863.

Lopes, N. M. (2025). Challenges and prospects of artificial intelligence in aviation: A bibliometric and systematic review. Journal of Air Transport Management, 128, 102054.

Mayer, M. (2023). Artificial intelligence and human-autonomy teaming in military systems. Journal of Defence Studies, 7(3), 45–61.

Molnar, T. G., Kousik, S., Singh, S., & Ames, A. D. (2024). Collision avoidance and geofencing for fixed-wing aircraft with control barrier functions. IEEE Transactions on Control Systems Technology, 32(5), 1954–1967.

Rashid, A. B. (2023). Artificial intelligence in the military: An overview of capabilities and risks. Computational Intelligence and Neuroscience, 2023, 1–12.

Sachdev, A. K. (2021). Artificial intelligence in military aviation. Air Power Journal, 16(2), 1–18.

Tafur, C. L., Gómez, J. A. Y Martínez, P. (2025). Applications of artificial intelligence in air operations. Aerospace Science and Technology, 152, 109123.

September 13, 2025September 28, 2025

739: ACCON 25: SECURING TOMORROW’S AVIATION IN AN AI AND QUANTUM-DRIVEN WORLD

ACCON 2025 KEYNOTE ADDRESS

Shared my views on the subject.

The march of computing power from the mechanical Wright Flyer of 1903 to the AI-powered, quantum-enabled systems of today has revolutionised aviation by heightening efficiency, automation, and connectivity.

Artificial intelligence (AI) is well embedded in aircraft operations, while quantum computing (QC) is still experimental but set to change design and logistics.

But these developments bring with them profound safety and security threats, and need to be addressed with strong mitigation measures.

Development of Computing Power in Aircraft

Pre-Computing Period (1903–1950s): Mechanical and Analogue Systems

1903 (Wright Flyer). No computing; manual operation through mechanical linkages and simple analogue instruments (e.g., compass, altimeter). Pilots relied on visual indicators.

1930s–1940s. Early commercial aircraft (e.g., Douglas DC-3) employed analogue instruments and ground radio beacons for navigation, with no computational processing.

Late 1940s. Analogue computers, such as gyroscopic autopilots in fighter aircraft, employed vacuum tubes for simple stabilisation.

Early Digital Computing (1950s–1970s): Analogue to Digital Shift

1950s. Analogue computers had stabilised but were heavy and restrictive.

1960s. Transistors allowed for digital avionics in military aircraft (e.g., F-4 Phantom) for simple navigation and radar. Commercial aircraft (e.g., Boeing 707) were still analogue-dominated.

Late 1960s–1970s. Integrated circuits (ICs), which borrowed from the Apollo Guidance Computer, featured limited digital processing on aircraft such as the Concorde (1969) with analogue fly-by-wire.

Digital Revolution (1980s–1990s): Fly-by-Wire and Glass Cockpits

1980s. Microprocessors created digital fly-by-wire (FBW) for the Airbus A320 (1987), utilising redundant processors (e.g., Intel 8086) to eliminate mechanical controls.

Glass Cockpits. Aircraft such as the Boeing 767 (1982) combined flight, navigation, and engine information on CRT screens.

Flight Management Systems (FMS). In the 1980s (e.g., Honeywell FMS), these utilised 16-bit processors to automate fuel management and navigation, lessening pilot workload.

Advanced Computing (1990s–2010s): Integration and Automation

1990s. PowerPC processors and GPS navigation in aircraft such as the Boeing 777 (1995) improved autopilot, diagnostics, and navigation.

Integrated Modular Avionics (IMA). The Airbus A380 (2005) integrated functions into centralised processors, enhancing efficiency.

Safety Systems. TCAS and EGPWS employed 32-bit processors (about hundreds of MIPS) for real-time collision avoidance and terrain clearance.

Current Period (2010s–2025): High-Performance Computing and AI

2010s. Multi-core processors (e.g., Intel/ARM) in planes like the Boeing 787 and Airbus A350 provided real-time weather analysis, predictive maintenance, and flight optimisation.

ADS-B (2020). Mandatory GPS-based position broadcasting necessitated high processing for traffic management.

Integration with AI. In 2025, AI systems (e.g., DARPA’s ALIAS, Boeing’s Loyal Wingman) will digest terabytes of sensor data for predictive maintenance, anomaly detection, and semi-autonomous flight.

Connectivity. High-bandwidth onboard servers are used to handle operational and passenger data.

Future Trends (2025 and beyond)

Quantum Computing. Research looks into QC for air traffic control and aerodynamics, with potential by the 2030s.

Sustainable Systems. Electric/hybrid aeroplanes (e.g., Airbus E-Fan X) use advanced battery management with real-time computing.

Autonomous Flight. AI-based systems with GPU/TPU accelerators handle petabytes of information for complete autonomous flight.

Influences of AI and Quantum Computing on Civil Aviation

Artificial Intelligence (AI)

Predictive maintenance. AI enables the evaluation of sensor data to predict failures, which increases reliability and reduces expenses.

Flight optimisation. AI improves fuel consumption by 10% and emissions by optimising routes and managing outages.

Autonomous Flight and Pilot Support. AI autopilot and co-pilot technologies take care of standard functions and optimise emergency responses, decreasing pilot workload.

Airport Performance. AI optimises check-in, baggage handling, and air traffic control, enhancing passenger journeys.

Design Innovation. Generative AI shortens aerodynamic and material design cycles.

Market Expansion. The AI aviation market is expected to expand at a 22.6% CAGR by 2030.

Quantum Computing (QC)

Advanced Simulations. QC optimises computational fluid dynamics (CFD) and structural analysis for light, efficient airframes.

Operational Optimisation. QC optimises difficult logistics issues (e.g., routing, cargo loading), potentially saving billions.

Sustainable Aviation. QC simulates new materials and fuels for hybrid/electric propulsion.

Future Potential. NASA and Boeing studies suggest QC advantages by the 2030s, in spite of existing error-rate limitations.

Effects of Computing Progression

Efficiency. FMS and route optimisation save billions in fuel costs each year.

Automation. Automated takeoffs, landings, and cruise allow for a single pilot, or sometimes autonomous flight, particularly in the case of military aviation.

Maintenance. Predictive maintenance, which relies on AI for the analysis of data, helps to reduce costs and delays.

Connectivity. Global data routed in near real-time enhances operations and passenger services.

Safety. With redundancy and real-time analysis (for example, TCAS, EGPWS), the accident rate has decreased more than 80% since the late 1970s.

Key Air Force AI Applications

Autonomous Combat Drones and Loyal Wingmen: AI-controlled UAVs, such as the U.S. Skyborg, Russia’s Okhotnik-B, and India’s CATS Warrior, conduct autonomous targeting, reconnaissance, and electronic warfare. Loyal wingmen (e.g., Boeing’s MQ-28 Ghost Bat) assist manned aircraft, minimising dangers to pilots.

AI-Assisted Air Combat. AI systems, as indicated in DARPA’s AlphaDogfight Trials, outcompete human pilots in dogfights through quick decision-making and optimal tactics.

AI Co-Pilot Systems. AI helps pilots with instant threat analysis, flight route optimisation, and weapons control, as in the U.S. Air Force’s ACE program.

Predictive Maintenance and Logistics. Artificial intelligence systems like CBM+ allow for the prediction of equipment failure, which reduces downtime and optimises allocation of resources, leading to improved fleet readiness and lower costs.

Air Defence Systems. AI allows for improved target detection and target engagement in air defence systems like Israel’s Iron Dome and Russia’s S-500 systems, allowing for a faster response to threats that are detected.

Electronic Warfare (EW). AI jams hostile radar independently, learns about threats, and defends assets against cyber and electromagnetic attacks.

Mission Planning. AI processes battlefield information to produce optimal plans, dynamically realigns plans, and incorporates multi-source intelligence for data-driven decision-making.

Swarm Warfare. Swarms of drones controlled by AI overwhelm defences, perform ISR, and jamming, with nations such as the U.S., China, and India developing this capability.

Benefits.

Better Decision Making. AI manages sizeable amounts of data for real-time intelligence and speed of reaction.

Reduction in Pilot Workload. Automators allow pilots to engage in tactically focused functions versus technically focused functions.

Improvement in Combat Effect. AI and drones enhance targeting.

Reduction in Collateral Damage. UAVs fly missions with high risk, ultimately reducing civilian casualties.

Creating levels of logistics. Predictive maintenance continues to reduce both operational downtime and costs.

Challenges & Ethical Issues

Autonomy versus Control. Fully autonomous systems raise a question of who is responsible.

Cybersecurity and Operational Risk. AI systems can be hacked and/or manipulated.

Bias and Mistakes. Incorrect target identification may result in unwanted collateral civilian casualties.

International Arms Race. The Race for sophisticated AI weapons systems potentially destabilises international security.

Prospects in the Future

Greater Autonomy. UCAVs will function independently in high-risk operations.

Hypersonic Weapons. AI will improve missile accuracy and velocity.

Quantum Integration. Artificial intelligence and quantum computing will transform data processing used in predictive analytics and threat detection.

Counter-AI Warfare. Armed forces will devise methods for nullifying adversary AI capabilities.

Ethical Regulation. Strong guidelines must be put in place to deal with ethical and strategic issues.

Security and Safety Risks to Aviation

Security Risks

Data Poisoning and Adversarial Attacks: AI inputs can maliciously be manipulated and affect flight controls, navigation, or airport functionality.

System Vulnerabilities. Ageing infrastructure can be susceptible to AI-based cyberattacks (e.g., ADS-B hijacking) and needs strong firewalls and intrusion detection.

Generative AI Threats. AI might be used to create deceptive data or evade security.

Encryption Threats. QC algorithms (e.g., Shor’s) might compromise public-key cryptography (RSA, ECC), endangering data breaches or spoofed signals in avionics and communications.

Harvest Now, Decrypt Later. Threats may carry encrypted flight data for later decryption, compromising flight plans and military communications.

Complex Attack Surfaces. Multiple layers of interconnected networks and avionics increase threats that are capable of quantum attacks.

Safety Risks

Algorithmic Errors. AI bias or misinterpretation can lead to incorrect commands for autopilot or navigation decisions, resulting in accidents.

Over-Reliance. AI reliance may negatively impact pilot proficiency; however, in-flight analysis can strengthen safety.

Transparency. Black box AI channels pilot interpretation and overt truth.

Semi-Autonomous Systems. The likelihood of a failure of autonomous operations in rare cases is significant.

Simulation Errors. QC’s current error rates could lead to defective designs exposed via QC, and lead to unsafe airframes.

Cyber-Driven Safety Critical Hazards. Quantum cyberattacks may disrupt avionics and navigation, leading to failures and unsafe operations.

Navigation Upgrades. Quantum sensors could provide fixes for navigation, but have not been adopted universally.

Mitigation Strategies

Post-Quantum Cryptography (PQC). Shift to quantum-resistant algorithms (e.g., lattice-based cryptography) to protect avionics, communications, and air traffic control. NIST is developing PQC standards.

Quantum Key Distribution (QKD). Use QKD for unbreakable encryption in high-priority systems such as ADS-B.

Resilient AI Governance. Build explainable AI (XAI) frameworks, ongoing validation, and adversarial testing to make it transparent and minimise errors.

Redundant Systems. Keep classical backups to counteract AI or QC failures.

Regulatory Harmonisation. Enhance global aviation standards for AI and QC certification with a priority on safety, interoperability, and training of the workforce.

Security by Design. Implement quantum-resistant architectures, identity-first safeguarding (e.g., biometrics, zero trust), and layered cyber defence in avionics and communications.

Automated with Human in the Loop. Implement AI-enabled automation (such as SOAR) to enhance response time while leveraging a human in the process in order to limit escalation.

Cloud Resilience. We need to balance our distributed cloud configurations and our sovereignty needs, imparting trust with these secure and reliable practices.

Conclusions

The computing capacity of mechanical devices in 1903 transitioned to AI-driven quantum systems by 2025. This change has transformed and continues to transform airline operations, enabling unprecedented levels of safety, efficiency, automation, and connectivity. AI is playing an ever-expanding role by improving The Boeing Root Cause Analysis for Maintenance, efficiencies in flight planning, and improving passenger experiences, with a projected 22.6 % CAGR growth to 2030. Quantum computing is yet largely experimental, but it is expected to have significant impacts on the way we design and logistics in the 2030s. We must study our speed of evolution against the risk and governance required with these technologies. AI has vulnerabilities either as a function of adversarial attacks or software imperfections, while quantum computing has the potential to break our encryption and create weaknesses in avionics and data integrity. There are safety risks we need to contend with, including failures in algorithms and problems in design from quantum technologies. The risk controls in the aviation environment require support for the cybersecurity principles established within ACCON’25, also known as the National Aerospace Cybersecurity Strategic Plan. These controls include Post-Quantum Cryptography (PQC), Quantum Key Distribution (QKD), Explainable AI (XAI), Redundant systems, Security-by.