Artificial Intelligence as a Means of Preventing Aviation Terrorism: Intent Detection, Automated Profiling and the Lessons of Flydubai Flight FZ1073
Abstract
On 30 September 2026, Hamam al-Hammami, the first officer of flydubai flight FZ1073 from Dubai to Tel Aviv, tried to take control of the aircraft after attacking the captain. According to authorities in the United Arab Emirates (UAE), he planned to crash the plane at Ben Gurion Airport. This article uses the incident to explore a wider question: how far can artificial intelligence (AI) and large language models (LLMs) help security services move from looking for dangerous items, such as weapons and explosives, to identifying harmful intentions? It examines four types of technology: advertising intelligence (ADINT), which gathers information without entering a person's device; real-time support for hostage negotiations; automated profiling of large numbers of people; and contactless tests designed to detect harmful intentions among flight crew. The article also discusses scientific limits, including the rarity of terrorist attacks, the limited reliability of methods for detecting lies, and the risk of wrongly identifying innocent people as threats. It also considers legal and ethical rules, including restrictions under the European Union's Artificial Intelligence Act. The main argument is that aviation security has not fully learned one of the key lessons of the September 11 attacks: a person's harmful intentions, rather than the tools they possess, may be the most important sign of danger. AI tools could help address this gap, but they must be tested with real evidence and used under suitable legal oversight.
Keywords: artificial intelligence; aviation security; insider threat; hostile-intent detection; profiling; advertising intelligence (ADINT); privacy
1. Introduction
Since large language models became widely available, the full effect of AI on security and intelligence has remained unclear. The idea of "preventing crime before it happens," made popular by the film Minority Report [1], has been put into practice in limited and debated forms through predictive policing [2]. Intelligence agencies have used computers to analyze data for decades. However, today's LLMs and other AI systems can process much more information, do so more quickly, and draw conclusions from information at a higher level of complexity.
The incident on flydubai flight FZ1073 provides an extreme and timely example of why these abilities may matter. According to reports, the first officer attacked the captain, Smit Machchhar, with the cockpit crash axe and tried to take control of the plane. Passengers and crew entered the cockpit and overpowered him. Two relief pilots who were on board then landed the aircraft safely in Tabuk, Saudi Arabia [3]. The UAE Attorney General said that the first officer had planned to crash the aircraft at Ben Gurion Airport, admitted that the September 11 attacks had inspired him, and had chosen a career in civil aviation so he could carry out a similar attack [4]. Other reports said he intended to continue the flight as usual, obtain permission to approach Ben Gurion, and crash the plane into a building or another aircraft [5]. This article asks whether AI-based technologies could have interrupted his plan at an early stage, and how much they might have helped.
2. Advertising Intelligence (ADINT): Tracking and Identifying Targets Without Entering Their Devices
Spyware that secretly enters computers and mobile phones and remains there, including tools made by companies that develop offensive cyber technology, has faced strong public and legal criticism. This criticism followed reports that such tools had been used against journalists and human rights activists [6, 7]. In 2021, the U.S. Department of Commerce added NSO Group to its Entity List [8]. Against this background, intelligence agencies have become more interested in ways to collect information without entering a device or leaving traces on it, while still tracking people of interest to intelligence services.
The main approach is advertising intelligence (ADINT). It uses the systems and data behind digital advertising to gather information for intelligence purposes. In online advertising, especially real-time bidding (RTB), each time an advertisement is shown, information about the device and its user is sent to many parties involved in the advertising process. This information can include the device's advertising ID, location, IP address, the app being used, and details about the user's behavior. Data brokers collect, combine, and sell this information. It can be bought or analyzed without directly accessing the device. A report to the U.S. Director of National Intelligence concluded that commercially available information (CAI) of this kind now makes it possible to identify people, follow their movements, and infer details about them on a scale and at a level of detail never seen before [9]. The academic study that introduced the term ADINT showed that someone with a limited budget could use targeted advertising systems to find a particular person and track the apps they use without leaving any trace on the person's device [10]. Unlike tools that break into devices, ADINT does not require exploiting software weaknesses, installing code, or gaining physical access. As a result, an examination of the device for evidence of an intrusion would not reveal this type of tracking.
Advertising data can also be used to build psychological profiles. Several studies have found that digital traces, such as Facebook "likes," can help predict personality traits and sensitive personal information [11]. In some cases, computer-based predictions have been more accurate than judgments made by people who know the person [12]. Research has also found that these methods are used in practice to influence consumer behavior through targeted messages [13]. The same tools that advertisers use to understand audiences and track their behavior could therefore be used by governments to follow intelligence targets and identify possible threats, including early signs of radicalization. However, such uses raise serious legal questions, discussed in Section 7.
3. Support for Hostage Negotiation
Crisis and hostage negotiators use structured communication methods, such as the Behavioral Change Stairway Model (BCSM) [14], and study how conversations change over time [15]. Commercial AI tools can now help negotiators during a live incident in several ways. They can automatically turn speech into text and translate it, with high accuracy even in noisy settings and across many languages [16]. They can analyze the people involved, assess a speaker's current state, monitor the conversation continuously, and warn when tensions are rising. They can also suggest changes in strategy and other possible actions. Research in natural language processing has shown that learning systems can learn negotiation strategies from conversations [17]. However, there is still little research on how well these tools work against actual terrorists.
4. Automated Profiling and the Detection of “Lone Wolves”
Large language models make it possible to create automated profiles of thousands or even tens of thousands of people. Recent research has found that an LLM can estimate the personality traits of social media users from the text they post, without being specially trained for that task. Its accuracy can be similar to that of machine-learning models designed specifically for this purpose [18]. Earlier studies found language patterns linked to radical violence on social media [19]. Research on terrorists who act alone has also found that many attacks were preceded by statements or behaviors that other people could observe, including cases where the person revealed their intentions to someone else [20]. These findings provide a theoretical basis for trying to identify a potential lone attacker before an attack occurs. They also suggest that a high-risk profile might be created for people such as tourists or people living in other countries.
However, assessing the risk that one person will commit terrorism has a major scientific limitation: terrorist attacks are extremely rare. Because the starting rate is so low, even a fairly accurate system can wrongly flag many people as dangerous [21]. Privacy is another important concern. One argument is that people who share information about themselves on social networks may, knowingly or unknowingly, make it available not only to advertisers' machine-learning systems but also to intelligence agencies.
5. Detecting Intent Versus Detecting Means
One of the main lessons of the September 11 attacks is that someone with harmful intentions does not need a powerful weapon, such as a gun or an improvised explosive device (IED), to carry out a deadly attack. The hijackers used small knives and the aircraft themselves [22]. Even so, most airports and airlines still focus on finding dangerous items in passengers' baggage and on their bodies rather than identifying harmful intentions. The FZ1073 incident makes this lesson even clearer: the attacker did not need to smuggle anything onto the plane. He used emergency equipment that was already inside the cockpit [3, 4].
The incident also shows that flight crew members, especially pilots, can pose a threat as serious as passengers, or possibly a greater one, because they have direct access to the aircraft's controls. A review of suicides and murder-suicides involving aircraft documented dozens of cases in general and commercial aviation [23]. These include well-known incidents such as the crash of Germanwings flight 9525 in 2015 [24] and EgyptAir flight 990 in 1999 [25]. Some aircraft have disappeared with little evidence left behind, including MH370. Although pilot involvement has been suggested as a possibility, the official cause of that event has not been determined [26].
In addition to gathering intelligence about flight crew, AI-based tests are available that aim to identify harmful intentions among crew members before a flight. Earlier systems for analyzing behavior and detecting harmful intentions mainly used sensors that had to touch the person, such as galvanic skin response (GSR) sensors [27, 28]. In recent years, users of systems designed to screen for insider threats [29] have shown growing interest in contactless tests. These use technologies such as remote photoplethysmography (rPPG) [30] and automated interview kiosks equipped with behavioral sensors [31]. Industry experience suggests that several requests for information (RFIs) about contactless intent detection have been issued in recent years. However, the main obstacle is not technology. It is concern about privacy and the possibility that the public and aviation labor unions will see such screening as intrusive, even when no physical contact is involved. After the Germanwings disaster, the European Union required operators to introduce peer-support programs and alcohol and drug testing for flight crew [32]. The area is therefore already regulated, but the existing rules focus on mental well-being rather than on identifying harmful intentions.
6. Case Study: How the FZ1073 Attack Could Have Been Disrupted
The following theoretical model combines different AI capabilities to examine whether the attack could have been prevented:
Intelligence collection and risk-profile building
- Monitor pilots around the world using ADINT, advertising data, commercially available information, and public sources such as social networks [9, 10].
- Build a personality profile and an updated risk profile for every flight crew member, including a risk score based on the threat being assessed [11, 18].
- Update each profile continuously, in real time.
- Look for signs of radicalization, visits to hostile states, or unusual assignments to flights to Israel, without entering the crew member's mobile device.
Intent detection
- Give crew members a pre-flight test for harmful intentions through a mobile app. A test lasting about three minutes would produce a list of crew members who need further background checks before they are allowed to fly.
It seems likely that a network of these systems would have flagged al-Hammami. According to reports, images related to terrorism were found on his social media accounts [33]. Omani authorities had previously stopped him from flying because of concerns about his extremist views, and he had been removed from flight training for similar reasons [3]. The UAE Attorney General said that he had been exposed to extremist material online since 2017 and had operated two earlier flights to Tel Aviv to become familiar with the route and airport [4]. These reports suggest that warning signs were present but were not shared or combined into one overall assessment of the risk. This is exactly the kind of gap that AI systems are intended to address. In addition, simply knowing that pilots and other crew members are being monitored by such systems might discourage similar attacks. This possibility is consistent with research on deterrence, which highlights the importance of how likely people believe they are to be caught [34].
7. Limitations and Ethical and Legal Considerations
A balanced discussion must consider several limitations. First, the scientific reliability of methods for detecting intentions and lies cannot be assumed. A report by the U.S. Government Accountability Office (GAO) found that the Transportation Security Administration's (TSA) behavior detection program had not been shown to work effectively and recommended limiting its funding [35]. The U.S. National Research Council found major weaknesses in the reliability of polygraph tests [27]. Reviews of research have also found that nonverbal signals are weak indicators of deception [36, 37]. Any AI system used for these purposes therefore needs independent testing with real-world evidence, clear reporting of error rates, and human judgment as part of the decision process. Second, as explained above, because this type of threat is rare, even a system that performs well can wrongly identify many people as threats [21]. Such mistakes can have serious consequences for people's jobs and personal lives.
Third, laws restrict some of the proposed uses. The European Union's Artificial Intelligence Act prohibits certain uses of emotion-recognition systems in the workplace, subject to exceptions. It also prohibits assessing whether someone may commit a crime based only on profiling, subject to the conditions set out in the law [38]. The General Data Protection Regulation (GDPR) restricts decisions based only on automated processing when those decisions have significant effects on people [39]. Using these tools to screen flight crew would therefore require a clear legal basis, proper oversight, and safeguards to ensure that a human makes the final decision.
8. Conclusion
The FZ1073 incident highlights the gap between security systems that look for dangerous items and the real threat posed by harmful intentions. AI technologies, including advertising intelligence (ADINT), automated profiling, and contactless tests for detecting harmful intentions, could help close this gap. They may be particularly useful for screening flight crew, a group that has received relatively little attention in this area. At the same time, privacy concerns, the views of labor unions, and legal restrictions may prevent or delay the use of these tools. Their scientific reliability must also be demonstrated through evidence. The issue brings one of the central conflicts of our time into focus: how to balance privacy and control over personal information with the need for security.
References
[1] Spielberg, S. (Director). (2002). Minority Report [Film]. 20th Century Fox; DreamWorks Pictures. (Based on P. K. Dick, “The Minority Report,” 1956.)
[2] Perry, W. L., McInnis, B., Price, C. C., Smith, S. C., & Hollywood, J. S. (2013). Predictive policing: The role of crime forecasting in law enforcement operations (RR-233). RAND Corporation.
[3] Al Jazeera. (2026, October 3). Who is the mystery co-pilot behind the Flydubai attack? [https://www.aljazeera.com/news/2026/10/3/who-is-the-mystery-co-pilot-behind-the-flydubai-attack](https://www.aljazeera.com/news/2026/10/3/who-is-the-mystery-co-pilot-behind-the-flydubai-attack)
[4] Israel National News. (2026, October 9). UAE: Flydubai co-pilot planned 9/11-style attack on Ben Gurion Airport. [https://www.israelnationalnews.com/news/434369](https://www.israelnationalnews.com/news/434369)
[5] Fox News. (2026, October 5). FlyDubai co-pilot allegedly planned to crash Israel-bound jet at Ben Gurion Airport. [https://www.foxnews.com/world/flydubai-copilot-allegedly-planned-crash-israel-bound-jet-ben-gurion-airport](https://www.foxnews.com/world/flydubai-copilot-allegedly-planned-crash-israel-bound-jet-ben-gurion-airport)
[6] Marczak, B., Scott-Railton, J., McKune, S., Abdul Razzak, B., & Deibert, R. (2018). Hide and seek: Tracking NSO Group’s Pegasus spyware to operations in 45 countries (Citizen Lab Research Report No. 113). University of Toronto.
[7] Amnesty International. (2021). Forensic methodology report: How to catch NSO Group’s Pegasus (DOC 10/4487/2021).
[8] U.S. Department of Commerce, Bureau of Industry and Security. (2021, November 3). Addition of certain entities to the Entity List. Federal Register, 86, 60759.
[9] Office of the Director of National Intelligence. (2022). Senior Advisory Group panel on commercially available information: Report to the Director of National Intelligence (declassified June 2023).
[10] Vines, P., Roesner, F., & Kohno, T. (2017). Exploring ADINT: Using ad targeting for surveillance on a budget — or — how Alice can buy ads to track Bob. Proceedings of the 2017 Workshop on Privacy in the Electronic Society (WPES ’17), 153–164.
[11] Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Sciences, 110(15), 5802–5805.
[12] Youyou, W., Kosinski, M., & Stillwell, D. (2015). Computer-based personality judgments are more accurate than those made by humans. Proceedings of the National Academy of Sciences, 112(4), 1036–1040.
[13] Matz, S. C., Kosinski, M., Nave, G., & Stillwell, D. J. (2017). Psychological targeting as an effective approach to digital mass persuasion. Proceedings of the National Academy of Sciences, 114(48), 12714–12719.
[14] Vecchi, G. M., Van Hasselt, V. B., & Romano, S. J. (2005). Crisis (hostage) negotiation: Current strategies and issues in high-risk conflict resolution. Aggression and Violent Behavior, 10(5), 533–551.
[15] Taylor, P. J. (2002). A cylindrical model of communication behavior in crisis negotiations. Human Communication Research, 28(1), 7–48.
[16] Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2023). Robust speech recognition via large-scale weak supervision. Proceedings of the 40th International Conference on Machine Learning (ICML), PMLR 202, 28492–28518.
[17] Lewis, M., Yarats, D., Dauphin, Y. N., Parikh, D., & Batra, D. (2017). Deal or no deal? End-to-end learning for negotiation dialogues. Proceedings of EMNLP 2017, 2443–2453.
[18] Peters, H., & Matz, S. C. (2024). Large language models can infer psychological dispositions of social media users. PNAS Nexus, 3(6), pgae231.
[19] Cohen, K., Johansson, F., Kaati, L., & Mork, J. C. (2014). Detecting linguistic markers for radical violence in social media. Terrorism and Political Violence, 26(1), 246–256.
[20] Gill, P., Horgan, J., & Deckert, P. (2014). Bombing alone: Tracing the motivations and antecedent behaviors of lone-actor terrorists. Journal of Forensic Sciences, 59(2), 425–435.
[21] Monahan, J. (2012). The individual risk assessment of terrorism. Psychology, Public Policy, and Law, 18(2), 167–205.
[22] National Commission on Terrorist Attacks upon the United States. (2004). The 9/11 Commission report. W. W. Norton.
[23] Kenedi, C., Friedman, S. H., Watson, D., & Preitner, C. (2016). Suicide and murder-suicide involving aircraft. Aerospace Medicine and Human Performance, 87(4), 388–396.
[24] Bureau d’Enquêtes et d’Analyses (BEA). (2016). Final report: Accident on 24 March 2015 at Prads-Haute-Bléone, Airbus A320-211, D-AIPX, operated by Germanwings.
[25] National Transportation Safety Board. (2002). Aircraft accident brief: EgyptAir Flight 990, Boeing 767-366ER, SU-GAP (NTSB/AAB-02/01).
[26] Australian Transport Safety Bureau. (2017). The operational search for MH370 (AE-2014-054).
[27] National Research Council. (2003). The polygraph and lie detection. The National Academies Press.
[28] Boucsein, W. (2012). Electrodermal activity (2nd ed.). Springer.
[29] Cappelli, D. M., Moore, A. P., & Trzeciak, R. F. (2012). The CERT guide to insider threats. Addison-Wesley.
[30] Verkruysse, W., Svaasand, L. O., & Nelson, J. S. (2008). Remote plethysmographic imaging using ambient light. Optics Express, 16(26), 21434–21445.
[31] Nunamaker, J. F., Jr., Derrick, D. C., Elkins, A. C., Burgoon, J. K., & Patton, M. W. (2011). Embodied conversational agent-based kiosk for automated interviewing. Journal of Management Information Systems, 28(1), 17–48.
[32] European Union. (2018). Commission Regulation (EU) 2018/1042 of 23 July 2018 amending Regulation (EU) No 965/2012 (support programmes, psychological assessment of flight crew, and testing for psychoactive substances). Official Journal of the European Union, L 188.
[33] CNN. (2026, October 2). Terrorist images found on Flydubai suspect’s social media. [https://www.cnn.com/2026/10/02/world/flydubai-attacker-pilot-intl](https://www.cnn.com/2026/10/02/world/flydubai-attacker-pilot-intl)
[34] Nagin, D. S. (2013). Deterrence in the twenty-first century. Crime and Justice, 42(1), 199–263.
[35] U.S. Government Accountability Office. (2013). Aviation security: TSA should limit future funding for behavior detection activities (GAO-14-159).
[36] Vrij, A., Granhag, P. A., & Porter, S. (2010). Pitfalls and opportunities in nonverbal and verbal lie detection. Psychological Science in the Public Interest, 11(3), 89–121.
[37] Denault, V., Plusquellec, P., Jupe, L. M., St-Yves, M., Dunbar, N. E., Hartwig, M., et al. (2020). The analysis of nonverbal communication: The dangers of pseudoscience in security and justice contexts. Anuario de Psicología Jurídica, 30, 1–12.
[38] European Union. (2024). Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 5(1)(d) and 5(1)(f). Official Journal of the European Union, L series.
[39] European Union. (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Article 22. Official Journal of the European Union, L 119.