When the Safety System Begins to Fail
Full Text / Article Transcript
When the Safety System Begins to Fail
Albert N. Clark
Independent Author
Published: September 1, 2026
ASX Research Journal and Database
ISSN 3068-3351 (Online)
Place of Publication: Cadiz City, Philippines
Publisher: ASXResearch.org
Author Note
Albert N. Clark
Department of Aerospace Sciences, ASXResearch.org
ORCID iD: https://orcid.org/0009-0002-7348-4395
The author reports no conflicts of interest.
Correspondence concerning this article should be addressed to Albert N. Clark, Email: [email protected]
Abstract
This article examines how aviation accidents can emerge from the progressive deterioration of organizational and institutional defenses rather than isolated human error. Drawing on the development of crew resource management, safety culture, quality assurance, voluntary reporting, and Safety Management Systems, the analysis explores how regulators, operators, manufacturers, military commands, training and maintenance organizations, airports, and investigative agencies form aviation’s multilayered safety architecture. Particular attention is given to normalization of deviance, staffing and experience pressures, weakened supervision, reporting culture, training deficiencies, organizational drift, and the distinction between regulatory compliance and operational resilience. Contemporary initiatives involving the Federal Aviation Administration, National Transportation Safety Board, manufacturers, the U.S. military, and DARPA are examined alongside the potential of artificial intelligence to identify institutional degradation before it enters an accident sequence. Legal and civil consequences are also considered when organizations recognize hazards but fail to act. The article concludes that the most dangerous safety system may be one that appears functional while progressively losing its ability to recognize and arrest deteriorating conditions. Future aviation safety depends upon detecting organizational drift, preserving institutional memory, protecting professional challenge, and using artificial intelligence to strengthen—not replace—the human and organizational defenses upon which aviation safety depends.
Keywords: aviation safety, organizational resilience, normalization of deviance
When the Safety System Begins to Fail
Aviation safety did not become reliable because pilots, mechanics, controllers, dispatchers, managers, or regulators suddenly stopped making mistakes; it became reliable because the industry gradually learned to surround fallible people with layers of institutional defense. Checklists, standardized training, maintenance inspection, independent oversight, voluntary reporting, quality assurance, recurrent qualification, accident investigation, and eventually Safety Management Systems were all built around the same uncomfortable premise: human error is inevitable, but catastrophe does not have to be. Crew resource management was especially important because it attacked the old hierarchical cockpit culture in which authority could suppress challenge and bad decisions could survive simply because the captain made them. Terzioğlu (2024) found that modern CRM remains significantly associated with flight safety culture, while Teperi et al. (2023) showed that human-factors programs become most effective when organizations move beyond regulatory compliance and transform them from an imposed requirement into an internalized professional mindset. The historical achievement of aviation safety, therefore, was not the elimination of human weakness; it was the creation of organizations expected to detect, challenge, contain, and learn from it before a single error could become an accident.
Figure 1
When the Safety System Begins to Fail: The Progressive Erosion of Organizational Defenses, Normalization of Deviance, and the Path Toward Systemic Aviation Failure
Note. Click image for full-size image or click here.
Safety Management Systems represented the institutionalization of that philosophy. Rather than waiting for wreckage to reveal hazards, SMS was designed to make organizations continuously identify risk, evaluate controls, monitor performance, encourage reporting, and modify operations before losses occurred. The weakness in that elegant architecture is that every component ultimately depends upon organizational behavior. A reporting system is useless if employees fear retaliation; a risk register is decorative if management accepts hazards indefinitely; recurrent training becomes theater if checking standards drift downward; and safety assurance is meaningless when the organization repeatedly verifies compliance without determining whether the underlying control actually works. Adjekum et al. (2023) demonstrated why generative voluntary reporting cultures are valuable precisely because frontline personnel often possess information about emerging hazards that formal organizational structures have not yet recognized. Key et al. (2023), working specifically with aviation safety culture, similarly found that safety culture can be measured and that it predicts important individual and organizational outcomes. These findings expose the central question of this paper: when credible warning information exists, why does an aviation organization sometimes fail to act on it?
One answer is normalization of deviance, a process far more dangerous than simple rule-breaking because it converts abnormal conditions into accepted reality. A shortcut works once, then again; a staffing deficit causes no accident; an inspection discrepancy proves harmless; a runway conflict ends with adequate separation; a controller manages an excessive workload without consequence; a maintenance team improvises around an unavailable part or incomplete record; and each successful outcome quietly becomes evidence that the unsafe condition was apparently safe after all. Sedlar et al. (2023), in a systematic review of normalization of deviance across high-risk industries, found that the phenomenon develops through repeated acceptance of departures from expected standards when adverse consequences fail to occur. Aviation is particularly vulnerable because its overall accident rate is so low that organizations can accumulate thousands of successful operations while underlying defenses deteriorate. The resulting institutional trap is brutal: absence of an accident is mistaken for evidence of control effectiveness. The system begins teaching itself that warning signs are normal, and eventually the organization is no longer managing risk; it is managing its familiarity with risk.
Maintenance provides one of the clearest windows into this process because maintenance safety depends simultaneously on technical competence, documentation, supervision, quality assurance, scheduling pressure, staffing, communication, and willingness to stop work. Aktas and Kagnicioglu (2023) found that organizational and individual factors significantly influence the safety behavior of aircraft maintenance technicians, reinforcing the point that technician performance cannot be separated from the environment created around the technician. Truong and Lee (2025) likewise found that aviation-maintenance safety culture requires a multidimensional framework rather than simple regulatory compliance, while Tyagi et al. (2023) showed that learning from previous maintenance events is essential to effective safety management. The institutional danger appears when an organization has all the required mechanisms on paper but the mechanisms become disconnected from operations: inspection becomes a signature, quality assurance becomes a production checkpoint, recurrent training becomes a calendar requirement, and a shortage of experienced technicians is compensated for by distributing responsibility among increasingly junior personnel. None of those conditions necessarily produces an immediate accident. Together, however, they can hollow out the very defenses that maintenance regulation was designed to create.
The same problem exists in flight operations and training. An airline can satisfy licensing, checking, simulator, duty, and qualification requirements while simultaneously becoming less resilient if experience is disappearing faster than it is being transferred. Chan et al. (2025) found that pilots with different training backgrounds attribute causal factors in safety events differently, which has direct implications for how airlines interpret incidents and design safety interventions. That matters because an organization whose workforce is increasingly junior cannot assume that identical certificates represent identical judgment, threat recognition, or operational context. Culture complicates the problem further. Pratama and Caponecchia (2025), reviewing research on national culture and aviation safety, found meaningful relationships between cultural factors and communication, human error, interaction, and safety performance. Consequently, airlines and training organizations cannot simply install CRM, SMS, and reporting programs as generic administrative products and assume equivalent outcomes. If deference discourages challenge, if instructors are overloaded, if checking becomes predictable, if staffing shortages discourage removal of marginal personnel from line operations, or if managers treat training failures primarily as production problems, the system can remain formally compliant while its real safety margin contracts.
Air traffic control illustrates the institutional issue even more starkly because controller performance is inseparable from staffing, traffic design, procedures, supervisory decisions, technology, and workload. Zamarreño Suárez et al. (2024) documented the extensive relationship between controller workload and safety, emphasizing that workload is not merely an individual endurance problem but a systems problem influenced by traffic complexity and operational design. The January 29, 2025 collision over the Potomac River subsequently provided an unusually severe demonstration of what happens when multiple institutional defenses fail together. In its 2026 final report, the NTSB identified systemic problems including FAA helicopter-route design, failure to evaluate and respond adequately to safety data, overreliance on visual separation, workload problems associated with combined control positions, and Army deficiencies involving helicopter operations and training; critically, the investigation did not reduce the disaster to one incompetent controller or pilot. That distinction matters. An accident caused by one reckless person can often be addressed by removing the person. An accident emerging from route design, operational policy, supervision, workload, training, technology, and ignored precursor data indicts the architecture that was supposed to prevent individual mistakes from becoming fatal.
Manufacturers are now confronting the same organizational question, and Boeing provides the most visible contemporary example. Following the 737 MAX crises and the January 2024 Alaska Airlines door-plug separation, Boeing expanded its enterprise Safety Management System, created stronger reporting mechanisms, reorganized safety responsibilities, increased workforce training, and developed its Boeing Safety Intelligence platform, which uses machine learning and advanced modeling to identify product and fleet safety signals. The NTSB’s investigation of Alaska Airlines Flight 1282 nevertheless showed why institutional reform cannot be judged by program names alone: missing door-plug bolts, inadequate production documentation, training deficiencies, ineffective oversight, and weaknesses in FAA auditing combined into a failure that should have been detected long before the airplane entered service. Airbus is pursuing a parallel but less crisis-driven approach, explicitly integrating open reporting, safety promotion, training, and SMS across its commercial, military, and unmanned-aircraft activities. The lesson is not that one manufacturer is safe and another unsafe; it is that manufacturers are increasingly treating safety culture, data analysis, employee reporting, and organizational design as engineering inputs rather than administrative matters.
Military aviation faces an even harder version of the problem because its safety system must coexist with mission urgency, tactical secrecy, command authority, combat realism, equipment limitations, and acceptance of risks that civil aviation would never tolerate. Fogarty et al. (2018) demonstrated in military aviation that safety climate is shaped by the balance between operational demands and organizational resources; high demand becomes especially dangerous when personnel perceive inadequate support, control, or resources. The U.S. Army’s response after the Potomac collision illustrates institutional adaptation rather than simple blame: changes have included tighter control of helicopter operations around Washington, changes involving ADS-B use, revised training practices, separation of some routine training from sensitive missions, and broader surveillance-equipment upgrades. This is precisely what a functioning safety system should do after identifying a systemic weakness: alter equipment, procedures, training, supervision, and operational policy together. Yet military organizations also face the danger of corrective-action decay. A post-accident surge in attention may produce dramatic improvements for several years, after which turnover, mission pressure, budget cycles, and institutional memory can gradually recreate the conditions that reform was intended to eliminate.
DARPA is tackling a related problem, but it is important not to claim more than the evidence supports. DARPA is not running a public program whose stated mission is to repair deteriorating airline safety culture, FAA oversight, maintenance quality assurance, or aviation workforce standards. Its Air Combat Evolution program instead uses autonomous air combat as a demanding test environment for building justified human trust in artificial intelligence and for understanding complex human-machine collaboration. That work is enormously relevant to the future of institutional safety because AI systems could eventually monitor trends that humans and fragmented departments miss: recurring unstable approaches, abnormal maintenance deferrals, fatigue patterns, reporting suppression, controller workload, training-performance drift, precursor combinations, or deviations becoming statistically normal. But automation cannot cure a dishonest organization. If executives define inconvenient data out of existence, supervisors discourage reports, commanders normalize risk, or regulators lack resources, the most sophisticated predictive system in the world becomes another dashboard to ignore. Elon Musk is even further removed from this particular problem. SpaceX is deeply engaged with federal aviation regulation and autonomous aerospace technology, but there is no credible public evidence that Musk, SpaceX, Tesla, or xAI is presently leading a program specifically aimed at correcting institutional degradation within airline, airport, maintenance, ATC, or military aviation safety systems. Claims that he is solving this particular problem would therefore outrun the evidence.
The FAA and NTSB are much more directly involved, although their roles differ profoundly. The FAA regulates and operates portions of the system and has progressively expanded mandatory SMS requirements; its 2024 rule broadened Part 5 SMS obligations to additional certificate holders and aviation organizations, while the 2023 airport SMS rule was explicitly intended to move certificated airports from traditional compliance toward proactive and eventually predictive risk management. The NTSB does not regulate; it investigates, identifies systemic failures, and recommends changes, and for years it has pushed broader and more effective implementation of SMS because nominal existence is not the same as verified effectiveness. Local, county, and state governments also occupy a larger role than is sometimes recognized because many public-use airports are owned or operated by cities, counties, airport authorities, or states. Under Part 139, those public entities may carry direct responsibility for runway safety, aircraft rescue and firefighting, wildlife control, fueling safety, markings, construction hazards, emergency planning, and—in covered airports—SMS implementation. Thus institutional aviation safety is not a federal pyramid with Washington at the top; it is a network in which a county airport operations office, municipal fire department, state aviation agency, airline safety department, manufacturer, military unit, and federal regulator may all control different pieces of the same accident pathway.
The legal consequences of institutional failure are equally systemic. When evidence shows that an organization knew or should have known of a recurring hazard yet failed to act, litigation can move beyond the frontline employee toward negligent hiring, training, supervision, retention, maintenance, operational control, product design, failure to warn, or organizational negligence, depending on the jurisdiction and facts. Airlines, manufacturers, maintenance providers, airport operators, contractors, and other private entities may face wrongful-death and personal-injury claims, while claims involving the FAA or other federal agencies implicate the Federal Tort Claims Act and its exceptions; military involvement creates additional statutory and sovereign-immunity complications, and state or municipal airport defendants may have separate governmental-immunity protections that vary by jurisdiction. Civil discovery can be devastating because internal emails, hazard reports, staffing records, audit findings, training failures, voluntary reports where legally discoverable, quality data, risk assessments, and management discussions may establish not merely that a hazard existed but that the organization had been warned. Criminal liability is far rarer and requires a substantially different evidentiary threshold, but deliberate falsification, obstruction, reckless conduct, or knowing regulatory violations can transform a safety breakdown into an enforcement or criminal matter. For safety leaders, the implication should not be “document less”; it should be the opposite: identify hazards candidly, document decisions intelligently, assign responsibility, verify mitigations, and never allow the written safety system to become evidence that management knew what needed to be done and chose not to do it.
The most important future development will therefore not be another layer of regulation by itself, another mandatory training module, or another artificial-intelligence product. It will be the transition from compliance-based safety management to continuously verified institutional resilience. The next generation of aviation safety systems should be capable of detecting not merely violations but drift: declining report rates, increasing training repeats, instructor shortages, maintenance rework, deferred discrepancies, overtime patterns, unstable staffing, recurring procedural workarounds, abnormal controller workload, weak supervisory intervention, and clusters of events that appear harmless individually but dangerous collectively. The central finding of this analysis is more troubling than the claim that aviation personnel are simply becoming less competent. It is that human performance can deteriorate in plain sight while every surrounding institution remains technically functional. The regulator still regulates, the airline still trains, the military still qualifies, the manufacturer still audits, the maintenance department still signs the paperwork, the airport still conducts inspections, and the SMS still produces reports—yet the defenses can become progressively less capable of stopping the accident sequence. That is when the safety system begins to fail: not when it disappears, but when it continues to exist while losing the courage, competence, authority, information flow, and institutional memory required to say no. The future role of AI should be to make that erosion harder to hide, not to provide aviation with a technological excuse for allowing it to continue.
References
Adjekum, D. K., Afari, S. A., Owusu-Amponsah, N. Y., Waller, Z., Rasouli, V., Ullrich, G., Snyder, P., & Corbin, N. (2023). Development of a generative voluntary safety reporting culture (GVSRC) model for the Gulf of Mexico (GOM) oil and gas (O & G) sector using attributes of the Aviation Safety Action Program (ASAP). Safety Science, 161, 106073. https://doi.org/10.1016/j.ssci.2023.106073
Aktas, E., & Kagnicioglu, C. H. (2023). Factors affecting safety behaviors of aircraft maintenance technicians: A study on civil aviation industry in Turkey. Safety Science, 164, 106146. https://doi.org/10.1016/j.ssci.2023.106146
Chan, W. T.-K., Li, W.-C., & Braithwaite, G. (2025). Pilots’ training backgrounds affecting the attribution of event causal factors and airline safety management. Journal of Air Transport Management, 125, 102786. https://doi.org/10.1016/j.jairtraman.2025.102786
Fogarty, G. J., Cooper, R., & McMahon, S. (2018). A demands–resources view of safety climate in military aviation. Aviation Psychology and Applied Human Factors, 8(2), 76–85. https://doi.org/10.1027/2192-0923/a000141
Key, K. N., Choi, I., Hu, P. T., & Schroeder, D. J. (2023). Validating a safety culture toolkit for aviation. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 67(1), 2030–2034. https://doi.org/10.1177/21695067231192404
Pratama, G. B., & Caponecchia, C. (2025). Examining the influence of national culture on aviation safety: A systematic review. Journal of Safety Research, 92, 317–330. https://doi.org/10.1016/j.jsr.2024.11.020
Sedlar, N., Irwin, A., Martin, D., & Roberts, R. (2023). A qualitative systematic review on the application of normalization of deviance within high-risk industries. Journal of Safety Research, 84, 290–305. https://doi.org/10.1016/j.jsr.2022.11.005
Teperi, A.-M., Paajanen, T., Asikainen, I., & Lantto, E. (2023). From must to mindset: Outcomes of human factor practices in aviation and railway companies. Safety Science, 158, 105968. https://doi.org/10.1016/j.ssci.2022.105968
Terzioğlu, M. (2024). The effects of crew resource management on flight safety culture: Corporate crew resource management (CRM 7.0). The Aeronautical Journal, 128(1326), 1743–1766. https://doi.org/10.1017/aer.2023.113
Truong, D., & Lee, S. A. (2025). Developing a new safety culture framework for aviation maintenance: Preliminary results. Safety Science, 183, 106729. https://doi.org/10.1016/j.ssci.2024.106729
Tyagi, A., Tripathi, R., & Bouarfa, S. (2023). Learning from past in the aircraft maintenance industry: An empirical evaluation in the safety management framework. Heliyon, 9(11), e21620. https://doi.org/10.1016/j.heliyon.2023.e21620
Zamarreño Suárez, M., Arnaldo Valdés, R. M., Pérez Moreno, F., Delgado-Aguilera Jurado, R., López de Frutos, P. M., & Gómez Comendador, V. F. (2024). Understanding the research on air traffic controller workload and its implications for safety: A science mapping-based analysis. Safety Science, 176, 106545. https://doi.org/10.1016/j.ssci.2024.106545