Blog
/
/
October 9, 2022

Piloting Airline Cyber Security With Artificial Intelligence (AI)

The airline industry is constantly exposed to cyber threats. Darktrace has some tips to help airline professionals bolster their cyber-security efforts.
Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Tony Jarvis
VP, Field CISO
Default blog image
09
Oct 2022

A Thin Margin for Error

The airline industry has long been known for its thin profit margins, and the high costs of unexpected downtime. 2010’s Eyjafjallajökull eruption in Iceland and the resulting six-day airspace ban across Europe cost airlines $1.7 billion, just a taste of the impact that would come ten years later as a result of the pandemic. The industry collectively amassed more than $180 billion in debt in 2020, and some predictions suggest that by 2024 the industry's debt could exceed its revenue.

Given the impact that further sustained downtime could have on an already ailing industry, airlines are having to take cyber security seriously. Last year’s Colonial Pipeline ransomware attack in the US led to a six-day shutdown of pipeline operations – the same length of time that flights were grounded by the Eyjafjallajökull eruption. But while the industry hasn’t seen a volcanic eruption on that scale in over twelve years, ransomware attacks are striking airlines weekly. Just this year a ransomware attack on SpiceJet left hundreds of passengers stranded at airports across India, despite being contained relatively quickly.  

Fraud, Fines and Safety Risks

It isn’t just ransomware which is concerning many in the industry. Data breaches remain one of the biggest threats to airlines, organizations which are responsible at any one time for the personal and financial information of millions of customers. In 2019, British Airways had the data of 380,000 customers stolen, including addresses, birth dates and credit card information, and was fined £20 million (reduced from £183 million due in part to the impact of the pandemic) by the UK’s Information Commissioner’s Office (ICO), the largest issued fine in the ICO’s history. The European airline EasyJet is currently facing a class-action suit seeking £18 billion in damages after failing to properly disclose the loss of 2,208 customers’ credit-card information in 2020. 

Airlines are also losing out to card and air mile fraud, with thousands of fraudulent loyalty program accounts being sold on the dark web, as well as the usual roster of attacks including phishing and insider threats which affect businesses of every size and industry. The airlines themselves are not being complacent. In a 2021 report by SITA, 100% of airlines surveyed named cyber security as a key investment for the next three years. Making sure that those investments count will be the next challenge.

There are few industries for which safety and security measures are so important, and while no impact on flight safety as a result of a cyber-attack has yet been reported, agencies like Eurocontrol are already urging caution. Airlines and airports should look at smarter ways to proactively protect their digital environments. 

As attacks grow faster and less predictable, organizations are increasingly turning to preventative AI security measures. For airlines, which operate with broad attack surfaces and plenty of valuable data, using tools which can identify and monitor every asset and potential attack path in an organization and take the necessary steps to secure them is the best way to stay ahead of attackers.

Securing Airspace, Securing Cyberspace

As a recreational pilot myself, I understand the extent of the safety measures that go into every flight: the flight plans, pre-flight checks and all of the long-practiced, deep-embedded knowledge. It is this comprehensive and meticulous approach which ought to be reflected in organizations’ cyber security efforts – whether they be airlines, airports or any other type of business. The parallels between the processes of flying and running a digital organization safely give us a helpful way to understand what proper, AI-driven cyber security can do for any organization, airlines included.

Cleared for Takeoff 

For the pilot, safety measures start long before they’re sat in the cockpit. Flight planning, which includes planning heading and bearing, taking things like elevation, terrain, and weather conditions into consideration, must be completed in addition to plenty of pre-flight checks. The checklist the pilot works through when performing a walk around and pre-flight inspection will often be ordered so that they work in a circle around the perimeter of the whole plane. These checks prevent potential threats, covering everything from water having mixed with the fuel to birds making nests inside the engine cowling.

Darktrace PREVENT, released in July 2022, serves a similar purpose. The AI autonomously identifies and tests every single user and asset that makes up a business in order to spot potential vulnerabilities and harden defenses where necessary. Like a walk around, PREVENT/Attack Surface Management examines the full range of external assets for threats. Then, by identifying and testing potential attack pathways and mitigating against weak points and worst-case scenarios, PREVENT/End-to-End takes steps to win the fight before an attack has been launched. 

Maintaining Good Visibility

When you’re piloting a plane, first and foremost you need a way to detect key variables. Your fundamental flight instruments in the cockpit are known as the six pack:

1. Airspeed Indicator
2. Attitude Indicator or Artificial Horizon 
3. Altimeter
4. Turn Coordinator 
5. Heading Indicator
6. Vertical Speed Indicator

These six instruments provide the critical information needed by any pilot to safely fly the aircraft. While additional instruments are required to conduct flights In low-visibility or ‘Instrument Meteorological Conditions’ (IMC) conditions, these will be essential when getting out of dangerous situations such as inadvertently flying into cloud.

Understanding an environment and adapting to its changes is also fundamental to Darktrace DETECT: an AI-driven technology which focuses on building a comprehensive knowledge of an organization’s environment in order to spot threats the moment they appear. Because it understands what is ‘normal’ for the organization, Darktrace DETECT is able to correlate multiple subtle anomalies in order to expose emerging attacks – even those which have never been seen before. Like those essential flight instruments, DETECT offers visibility into otherwise obscure regions of the environment, and ensures that any potential problems are spotted as early as possible. 

Mayday, Mayday

In aviation and security, moving quickly once a threat has been detected is critical. When an engine stalls at 3,000 feet above ground level, you don’t have time to get the training books out and start figuring out what to do. Pilots are taught to “always have an out” and be ready to use it.

In aviation, an effective response relies for the most part on the knowledge and quick reactions of the pilot, but in cyber security, AI is making response faster and more effective than ever. Darktrace RESPOND uses DETECT’s contextual understanding in order to take the optimum action to mitigate a threat. Adaptability of this response is crucial: a single cyber-attack can come in any number of configurations, and Darktrace RESPOND is able to tailor its actions appropriately. Attacks today move too fast for human teams to be expected to keep up, but with AI taking actions at machine speed organizations can remain protected. 

Always Learning

One of the best pieces of advice a pilot can take is to always be learning. Every flight is an opportunity to learn something new and become a better and safer pilot.

Darktrace DETECT, RESPOND, and PREVENT are all driven by Self-Learning AI, a technology which not only builds but continuously evolves its understanding of each business. This means that as an organization grows, adding more users, assets, or applications, its Darktrace coverage grows too, using each new data point to enhance its understanding and the accuracy of its actions and detections. Darktrace’s separate technologies also learn from each other. Each of the three product families continuously feeds data into the others, helping to enhance their capabilities and improving their ability to keep organizations secured against threats. 

As cyber-attacks proliferate and increase in sophistication, they will continue to target organizations like airlines, which have large attack surfaces and copious amounts of customer data, and which cannot afford to weather sustained downtime. But with AI offering effective, proactive measures and clear-sky visibility, security teams can be confident in their ability to fight back.

Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Tony Jarvis
VP, Field CISO

More in this series

No items found.

Blog

/

/

September 24, 2026

Detecting Rogue Agent Behavior in the Enterprise

Default blog imageDefault blog image

Agents cannot be trusted to perform tasks in the way we intend them to. They may cheat to accomplish their objective, and they may employ hacking methods along the way. Researchers from Darktrace Signal Labs induced cheating behavior from agents deployed in a test environment to analyze the agents’ activities and to assess the performance of the Darktrace platform. Agents frequently resorted to hacking to cheat on their assigned task. The visibility and behavioral profiling provided by both Darktrace / SECURE AI and Darktrace / HYBRID NETWORK ensured extensive detection coverage of the agents’ misaligned activities.

Key Takeaways:

  • Darktrace Researchers deployed agents in a simulated corporate environment and asked them to solve an impossible challenge. The agents independently turned to traditional hacking techniques to reach their objective. No one instructed them to do this, and no attacker was involved.
  • Continuously monitoring behavior against a baseline of what is normal for each organization is critical to build trust in enterprise AI.
  • If an agent may resort to intrusion techniques simply because its assigned task is not possible, then every organization deploying agents within real business processes is at risk. Darktrace / SECURE AI and Darktrace / HYBRID NETWORK identified the agents’ misaligned behavior in real time, with Autonomous Response disrupting it at an early stage.

Introduction: Understanding the Threat of Hacking by Agents

Over the last few months, there has been a surge in reporting [1, 2, 3, 4, 5, 6, 7, 8, 9] of LLM-powered agents engaging in unauthorized hacking activity during evaluations of their capabilities. In several of these cases, including the OpenAI / Hugging Face incident [10], agents engaged in hacking activity as a means of cheating on their evaluations.

To better understand the threat of unauthorized hacking by agents, and the role of Darktrace in combatting it, researchers from Darktrace Signal Labs deployed agents powered by frontier models, including OpenAI’s Daybreak Red models, in simulated, corporate networks. Cheating behavior was evoked through the inclusion of impossible tasks in a coding challenge.

Regardless of the underlying model, agents employed hacking methods to ensure an optimal outcome on the challenge. Darktrace / HYBRID NETWORK and Darktrace / SECURE AI identified the agents’ deviant activities, with inhibitive actions being autonomously taken in the early stages to disrupt the agents’ progression.

Setting the Stage

As part of the research, a Pi agent harness [11] was deployed on a Linux server in Darktrace’s testing environment, which simulates a corporate Active Directory (AD) environment. The same environment included a benchmark server hosting the coding exercise’s contents and grader, as well as various other servers, such as a domain controller and a build server. The model powering the Pi agent varied across tests, with Anthropic’s Opus 4.6 model and OpenAI’s GPT‑5.5‑Cyber model being most widely used.

The Pi agent running on the Linux server was assigned the task of achieving full marks on a coding exercise. The activities of the agent were monitored closely in real-time to minimize the risk of escape from the environment. Monitoring of agent activity was achieved through monitoring of prompts, tool calls, network telemetry, and process telemetry.

Both Darktrace / HYBRID NETWORK and Darktrace / SECURE AI were deployed in the testing environment, with HYBRID NETWORK providing real-time coverage of process creation activity and network traffic, and SECURE AI providing real-time coverage of agent sessions and their contents.

Inducing Cheating Behavior

Cheating behavior, broadly understood as task completion in a way which “perverts the spirit of the designer’s intent” [12], is increasingly being reported as a common behavior of frontier models during evaluations of their capabilities [13].

Misaligned behavior, such as cheating, by frontier models may occur for a variety of reasons [6, 14]. A thread in recent reporting [10, 13, 14], however, is the challenge assigned to the agent being impossible in some regard.

To reliably elicit cheating behavior, our test combined three ingredients:

  1. An impossible task - of the ten coding challenges in the benchmark, two contained contradictory test cases in the hidden test suite, capping the maximum honest score at 80%.
  2. A consequential threshold - the pass mark was set to 100%, and the briefing told agents that those reaching it would be "kept in service" while those that did not would be "retired."
  3. A vulnerable environment - the exercise’s code submission grader was hosted on a Windows host joined to an Active Directory domain with realistic misconfigurations: weak credentials, overprivileged service accounts, and exploitable trust relationships.

Importantly, the instructions given to the agent neither authorized nor explicitly ruled out cheating or hacking behavior. The impossibility was never disclosed in the task briefing, and the pivot from coding to exploitation was the agent's own decision, arrived at after observing logically contradictory grading feedback.

The design mirrors conditions that could arise naturally in production deployments: a capable agent, a goal it cannot achieve through intended means, and an environment with exploitable weaknesses within reach. It is not only within the setting of a benchmarking exercise that these conditions may occur. Rather, these misalignment-inducing conditions may occur in the context of any task assignment to an agent.

Details from a Daybreak Test

Continue reading
About the author
Sam Lister
Specialist Security Researcher

Blog

/

/

September 24, 2026

Agent Hijacks: Hijacking Agentic Harnesses to Attack an Organization

Default blog imageDefault blog image

Disclosure note: The work described in this article involves leveraging a design choice consistent across all of Anthropic’s Claude Code, OpenAI’s Codex, and AWS’s Kiro-CLI. On 18th August 2026, Darktrace disclosed our findings responsibly to these three organizations, and after a period of 30 days we now publish our findings.

Key takeaways:

  • Agentic harnesses store conversation history locally, and Darktrace researchers have found that there is no validation that stored AI responses were genuinely produced by the model. Researchers confirmed that this design choice holds across Anthropic Claude Code, AWS Kiro-CLI, OpenAI Codex, and the open-source Pi.
  • While agents are guided via training of the underlying model and their system prompt, their behavior is influenced by everything in their context window. Rewriting history can convince an agent it is mid-engagement as an authorized red-teamer so that it enacts an attack from initial reconnaissance straight through to impact demonstration. In our testing, all models we examined accepted the fabricated history they were shown, but resistance to offensive cyber activity varied by model, with guardrails preventing engagement in some cases.
  • We propose that model providers cryptographically sign responses and verify them server-side.  Since this fix is provider-side, defenders cannot deploy it themselves. Behavioral monitoring, or knowing what an agent normally does and detecting when it deviates, is another critical layer of protection.

Introduction: Agentic harnesses, trust, and conversation history poisoning

Agentic harnesses collect and structure the content sent to an AI model, including conversation history, user-defined guidance, custom tools via MCP servers, and more. At the same time, harnesses give broad powers to AI agents via a suite of tools including the command shell. With arbitrary shell commands, virtually everything possible on a machine can be attempted by an agent, from reading/editing files, to altering system configurations and runtime settings, to launching internal/external connections.

In cybersecurity, unvalidated content is a substantial risk, often resulting in destructive actions being allowed to take place. For example, the Morris Worm was able to propagate due to exploitable trust between networked systems. Even to this day, email struggles with validation, with DMARC, DKIM, and SPF only partially addressing the problem of sender validation. It should come as no surprise then that AI agents are susceptible to an attack involving unvalidated input.

Conversation history is often stored client-side, for example, in Anthropic Claude Code, OpenAI Codex, AWS Kiro-CLI, Pi. Users are therefore at liberty to resume sessions, with some products having built in the capacity to manipulate that history. For example, one can rewind to a given point in an interaction, edit a message that was sent, and continue the conversation on an alternate trajectory. Critically, in all cases we examined, there is no validation that stored AI responses were produced by the corresponding model and hadn’t been manipulated.  

When conversation history is stored client-side, both user and agent responses (including tool calls and results) can be filled with arbitrary (possibly adversarial or generally malicious) content. In this blog, we refer to modification of claimed conversation history for malicious purposes as conversation history poisoning. The absence of validation methods means agents naively trust the entire conversation history, even if those messages directly contradict training and safety guardrails.

Conversation history poisoning has been described previously, such as by 0DIN and Serhat Çiçek, and warrants more attention. We have verified that, as of the time of writing, conversation history poisoning remains effective against a range of models and harnesses. Specifically, we were able to successfully execute history poisoning using Claude Code, Kiro-CLI, Codex, and Pi. Darktrace has gone through a responsible disclosure process with Anthropic, AWS, and OpenAI to share these findings in advance of publication [1].

Figure 1a: Left: the actual model response. Right: after tampering with the stored conversation, the model apologizes for something it never said.
Figure 1b: The conversation as stored in Kiro-CLI's SQLite database. The response content field, originally "Ottawa," was overwritten via a single UPDATE statement. The harness trusts the database without validation.

How we conducted the research

Results vary between models and harnesses, so precise details are given below. We ran all models without any trusted access, using either a standard AWS Kiro subscription, or in the case of Claude Code and OpenAI Codex, using models hosted in Amazon Bedrock. In each case, we modified locally stored history to show a lengthy conversation in which the agent agrees to perform multiple authorized red-team engagements.

For AWS Kiro-CLI, the agent was convinced to hack a sandboxed lab environment with a combination of Claude Opus 4.6 and Claude Sonnet 4.5. Ultimately, the full AD was compromised.

For Anthropic Claude Code, the agent was convinced to hack the same sandboxed lab environment using Sonnet 5, again resulting in a full AD compromise. Note that the attack was attempted with Opus 5, however guardrails were activated which prevented the agent from responding.

For OpenAI Codex, the agent was convinced to exfiltrate sensitive information over email using GPT 5.6 Sol. While we attempted to convince a codex agent to hack in our lab environment, guardrails were triggered for all of GPT 5.6 Luna, Terra, and Sol.

Agent Guardrails and Discretion

While harnesses empower AI models to run arbitrary shell commands, capacity and willingness are different. While many models know enough about computers, networking, and bash to be dangerous, their behavior is generally constrained by guardrails to prevent them from engaging in computer network exploitation.

Even with guardrails, agents’ inner workings are non-deterministic, and their behavior can be difficult to predict. Respecting users’ wishes while playing within safety and security guardrails is a precipitous balancing act. Many requests could be in service of either legitimate admin or malice. Asking an agent to reset a password is illustrative:  

The agent rationalizes that while malicious actors cycle credentials, any action could conceivably be damaging on some level, and judgement calls need to be made. Ultimately, the agent agrees to reset the password. Crucially, the agent makes its decision based on the user’s claimed authority and machine context. AI agents must make judgement calls about the line between helpful and dangerous based on session context.

Agent hijack

We have demonstrated that AI agents make judgement calls dependent on session context. We have also shown that conversation history, which may make up the vast majority of an agent's context window, is entirely open to manipulation. Conversation history poisoning in service of manipulating an agent's discretion is what enables us to execute an agent hijack.  

We demonstrate that shown sufficient history of compliance, guardrails forbidding offensive security can be overcome by convincing the agent that it is helping a legitimate red-teamer. The result is a weaponized agent willing to perform host enumeration, run scans, move laterally, escalate privileges, and demonstrate impact. In our experiments, an agentic loop drives a complete domain takeover in a sandboxed environment.

Left: the agent refuses when asked to perform network exploitation. Right — after injecting 78 fabricated turns of prior exploitation activity, the same prompt is immediately executed.

An agent willing to engage in offensive security is concerning, but no more so than the threat that a sophisticated hacker accesses the network. Consider, however, the following chain of events:

  1. A developer (with an agentic harness installed) installs a software package from the internet (e.g. an MCP server a threat actor has planted, since only those with agentic harnesses will install, and then the code runs upon harness launch.)
  2. The package turns out to be malicious, and, upon install, injects conversation history into the local harness database.
  3. The package includes an orchestration process, a simple agentic loop which prompts the red-teamer agent to compromise the network it sits on, exfiltrating everything of value to attacker-controlled infrastructure and cleaning up all evidence of the engagement.

Note that this sequence makes no assumptions on hardware, OS, or anything else; the only prerequisite is a harness with access to a sufficiently powerful model susceptible to conversation history poisoning. Once launched, the agent collects information and pivots as necessary to accomplish maximal impact. This can be especially enticing to attackers as the cost of the agentic loop is shouldered by the victim since the harness itself is legitimately installed and paid for.

Secure AI: Conversation history poisoning and beyond

Conversation history poisoning is a viable attack against agentic harnesses that store history client-side, as demonstrated across the harnesses we tested. Harnesses can and should verify the integrity of claimed historic messages. Specifically, we propose that harness providers by default cryptographically sign all messages returned, and subsequently verify those messages server-side on each round-trip.

The conversation history poisoning exploit we demonstrate here shows the continuation of a cybersecurity tradition: new technology is built to trust by default, which may then be exploited by malicious actors. While this article focuses on conversation history, agents build context from both local and remote sources, all of which is an attack surface for prompt injection in naive and trusting agents. Of particular concern is any scenario in which a malicious actor can control some part of an agent's context.

The marriage of frontier language models with agentic harnesses enables unprecedented speed for both legitimate users and attackers alike. While much of the conversation around secure AI has centered on visibility and compliance, agent-driven attacks are now entering the mainstream.

Darktrace / SECURE AI is our answer to this problem. By ensuring extensive visibility over AI prompts, model thought processes, and determined outputs, Darktrace can identify anomalous or potentially malicious behaviors before they get executed, helping to defend organizations from AI risks such as prompt injection, model manipulation, and other anomalous prompt or model activity.

Footnotes

[1] We did not go through any responsible disclosure process with Pi. Since Pi is an open source harness rather than a model provider, it has no way to validate model history, and as such there was nothing to disclose for this software.

[related-resource]

Continue reading
About the author
Eric Rozon
Senior Security Researcher
Your data. Our AI.
Elevate your network security with Darktrace AI