October 30, 2023

Exploring AI Threats: Package Hallucination Attacks

Learn how malicious actors exploit errors in generative AI tools to launch packet attacks. Read how Darktrace products detect and prevent these threats!

Written by

Charlotte Thompson

Cyber Analyst

Written by

Tiana Kelly

Deputy Team Lead, London & Cyber Analyst

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

Charlotte Thompson

Cyber Analyst

Written by

Tiana Kelly

Deputy Team Lead, London & Cyber Analyst

Oct 2023

AI tools open doors for threat actors

On November 30, 2022, the free conversational language generation model ChatGPT was launched by OpenAI, an artificial intelligence (AI) research and development company. The launch of ChatGPT was the culmination of development ongoing since 2018 and represented the latest innovation in the ongoing generative AI boom and made the use of generative AI tools accessible to the general population for the first time.

ChatGPT is estimated to currently have at least 100 million users, and in August 2023 the site reached 1.43 billion visits [1]. Darktrace data indicated that, as of March 2023, 74% of active customer environments have employees using generative AI tools in the workplace [2].

However, with new tools come new opportunities for threat actors to exploit and use them maliciously, expanding their arsenal.

Much consideration has been given to mitigating the impacts of the increased linguistic complexity in social engineering and phishing attacks resulting from generative AI tool use, with Darktrace observing a 135% increase in ‘novel social engineering attacks’ across thousands of active Darktrace/Email™ customers from January to February 2023, corresponding with the widespread adoption of ChatGPT and its peers [3].

Less overall consideration, however, has been given to impacts stemming from errors intrinsic to generative AI tools. One of these errors is AI hallucinations.

What is an AI hallucination?

AI “hallucination” is a term which refers to the predictive elements of generative AI and LLMs’ AI model gives an unexpected or factually incorrect response which does not align with its machine learning training data [4]. This differs from regular and intended behavior for an AI model, which should provide a response based on the data it was trained upon.

Why are AI hallucinations a problem?

Despite the term indicating it might be a rare phenomenon, hallucinations are far more likely than accurate or factual results as the AI models used in LLMs are merely predictive and focus on the most probable text or outcome, rather than factual accuracy.

Given the widespread use of generative AI tools in the workplace employees are becoming significantly more likely to encounter an AI hallucination. Furthermore, if these fabricated hallucination responses are taken at face value, they could cause significant issues for an organization.

Use of generative AI in software development

Software developers may use generative AI for recommendations on how to optimize their scripts or code, or to find packages to import into their code for various uses. Software developers may ask LLMs for recommendations on specific pieces of code or how to solve a specific problem, which will likely lead to a third-party package. It is possible that packages recommended by generative AI tools could represent AI hallucinations and the packages may not have been published, or, more accurately, the packages may not have been published prior to the date at which the training data for the model halts. If these hallucinations result in common suggestions of a non-existent package, and the developer copies the code snippet wholesale, this may leave the exchanges vulnerable to attack.

Research conducted by Vulcan revealed the prevalence of AI hallucinations when ChatGPT is asked questions related to coding. After sourcing a sample of commonly asked coding questions from Stack Overflow, a question-and-answer website for programmers, researchers queried ChatGPT (in the context of Node.js and Python) and reviewed its responses. In 20% of the responses provided by ChatGPT pertaining to Node.js at least one un-published package was included, whilst the figure sat at around 35% for Python [4].

Hallucinations can be unpredictable, but would-be attackers are able to find packages to create by asking generative AI tools generic questions and checking whether the suggested packages exist already. As such, attacks using this vector are unlikely to target specific organizations, instead posing more of a widespread threat to users of generative AI tools.

Malicious packages as attack vectors

Although AI hallucinations can be unpredictable, and responses given by generative AI tools may not always be consistent, malicious actors are able to discover AI hallucinations by adopting the approach used by Vulcan. This allows hallucinated packages to be used as attack vectors. Once a malicious actor has discovered a hallucination of an un-published package, they are able to create a package with the same name and include a malicious payload, before publishing it. This is known as a malicious package.

Malicious packages could also be recommended by generative AI tools in the form of pre-existing packages. A user may be recommended a package that had previously been confirmed to contain malicious content, or a package that is no longer maintained and, therefore, is more vulnerable to hijack by malicious actors.

In such scenarios it is not necessary to manipulate the training data (data poisoning) to achieve the desired outcome for the malicious actor, thus a complex and time-consuming attack phase can easily be bypassed.

An unsuspecting software developer may incorporate a malicious package into their code, rendering it harmful. Deployment of this code could then result in compromise and escalation into a full-blown cyber-attack.

Figure 1: Flow diagram depicting the initial stages of an AI Package Hallucination Attack.

For providers of Software-as-a-Service (SaaS) products, this attack vector may represent an even greater risk. Such organizations may have a higher proportion of employed software developers than other organizations of comparable size. A threat actor, therefore, could utilize this attack vector as part of a supply chain attack, whereby a malicious payload becomes incorporated into trusted software and is then distributed to multiple customers. This type of attack could have severe consequences including data loss, the downtime of critical systems, and reputational damage.

How could Darktrace detect an AI Package Hallucination Attack?

In June 2023, Darktrace introduced a range of DETECT™ and RESPOND™ models designed to identify the use of generative AI tools within customer environments, and to autonomously perform inhibitive actions in response to such detections. These models will trigger based on connections to endpoints associated with generative AI tools, as such, Darktrace’s detection of an AI Package Hallucination Attack would likely begin with the breaching of one of the following DETECT models:

Compliance / Anomalous Upload to Generative AI
Compliance / Beaconing to Rare Generative AI and Generative AI
Compliance / Generative AI

Should generative AI tool use not be permitted by an organization, the Darktrace RESPOND model ‘Antigena / Network / Compliance / Antigena Generative AI Block’ can be activated to autonomously block connections to endpoints associated with generative AI, thus preventing an AI Package Hallucination attack before it can take hold.

Once a malicious package has been recommended, it may be downloaded from GitHub, a platform and cloud-based service used to store and manage code. Darktrace DETECT is able to identify when a device has performed a download from an open-source repository such as GitHub using the following models:

Device / Anomalous GitHub Download
Device / Anomalous Script Download Followed By Additional Packages

Whatever goal the malicious package has been designed to fulfil will determine the next stages of the attack. Due to their highly flexible nature, AI package hallucinations could be used as an attack vector to deliver a large variety of different malware types.

As GitHub is a commonly used service by software developers and IT professionals alike, traditional security tools may not alert customer security teams to such GitHub downloads, meaning malicious downloads may go undetected. Darktrace’s anomaly-based approach to threat detection, however, enables it to recognize subtle deviations in a device’s pre-established pattern of life which may be indicative of an emerging attack.

Subsequent anomalous activity representing the possible progression of the kill chain as part of an AI Package Hallucination Attack could then trigger an Enhanced Monitoring model. Enhanced Monitoring models are high-fidelity indicators of potential malicious activity that are investigated by the Darktrace analyst team as part of the Proactive Threat Notification (PTN) service offered by the Darktrace Security Operation Center (SOC).

Conclusion

Employees are often considered the first line of defense in cyber security; this is particularly true in the face of an AI Package Hallucination Attack.

As the use of generative AI becomes more accessible and an increasingly prevalent tool in an attacker’s toolbox, organizations will benefit from implementing company-wide policies to define expectations surrounding the use of such tools. It is simple, yet critical, for example, for employees to fact check responses provided to them by generative AI tools. All packages recommended by generative AI should also be checked by reviewing non-generated data from either external third-party or internal sources. It is also good practice to adopt caution when downloading packages with very few downloads as it could indicate the package is untrustworthy or malicious.

As of September 2023, ChatGPT Plus and Enterprise users were able to use the tool to browse the internet, expanding the data ChatGPT can access beyond the previous training data cut-off of September 2021 [5]. This feature will be expanded to all users soon [6]. ChatGPT providing up-to-date responses could prompt the evolution of this attack vector, allowing attackers to publish malicious packages which could subsequently be recommended by ChatGPT.

It is inevitable that a greater embrace of AI tools in the workplace will be seen in the coming years as the AI technology advances and existing tools become less novel and more familiar. By fighting fire with fire, using AI technology to identify AI usage, Darktrace is uniquely placed to detect and take preventative action against malicious actors capitalizing on the AI boom.

Credit to Charlotte Thompson, Cyber Analyst, Tiana Kelly, Analyst Team Lead, London, Cyber Analyst

References

[1] https://seo.ai/blog/chatgpt-user-statistics-facts

[2] https://darktrace.com/news/darktrace-addresses-generative-ai-concerns

[3] https://darktrace.com/news/darktrace-email-defends-organizations-against-evolving-cyber-threat-landscape

[4] https://vulcan.io/blog/ai-hallucinations-package-risk?nab=1&utm_referrer=https%3A%2F%2Fwww.google.com%2F

[5] https://twitter.com/OpenAI/status/1707077710047216095

[6] https://www.reuters.com/technology/openai-says-chatgpt-can-now-browse-internet-2023-09-27/

Deputy Team Lead, London & Cyber Analyst

Inside the SOC

Written by

Charlotte Thompson

Cyber Analyst

Written by

Tiana Kelly

Deputy Team Lead, London & Cyber Analyst

Global Threat Research Operations Lead

Watch the NIS2 Webinar

Blog

AI

February 26, 2026

What the Darktrace Annual Threat Report 2026 Means for Security Leaders

The challenge for today’s CISOs

At the broadest level, the defining characteristic of cybersecurity in 2026 is the sheer pace of change shaping the environments we protect. Organizations are operating in ecosystems that are larger, more interconnected, and more automated than ever before – spanning cloud platforms, distributed identities, AI-driven systems, and continuous digital workflows.

The velocity of this expansion has outstripped the slower, predictable patterns security teams once relied on. What used to be a stable backdrop is now a living, shifting landscape where technology, risk, and business operations evolve simultaneously. From this vantage point, the central challenge for security leaders isn’t reacting to individual threats, but maintaining strategic control and clarity as the entire environment accelerates around them.

Strategic takeaways from the Annual Threat Report

The Darktrace Annual Threat Report 2026 reinforces a reality every CISO feels: the center of gravity isn’t the perimeter, vulnerability management, or malware, but trust abused via identity. For example, our analysis found that nearly 70% of incidents in the Americas region begin with stolen or misused accounts, reflecting the global shift toward identity‑led intrusions.

Mass adoption of AI agents, cloud-native applications, and machine decision-making means CISOs now oversee systems that act on their own. This creates an entirely new responsibility: ensuring those systems remain safe, predictable, and aligned to business intent, even under adversarial pressure.

Attackers increasingly exploit trust boundaries, not firewalls – leveraging cloud entitlements, SaaS identity transitions, supply-chain connectivity, and automation frameworks. The rise of non-human identities intensifies this: credentials, tokens, and agent permissions now form the backbone of operational risk.

Boards are now evaluating CISOs on business continuity, operational recovery, and whether AI systems and cloud workloads can fail safely without cascading or causing catastrophic impact.

In this environment, detection accuracy, autonomous response, and blast radius minimization matter far more than traditional control coverage or policy checklists.

Every organization will face setbacks; resilience is measured by how quickly security teams can rise, respond, and resume momentum. In 2026, success will belong to those that adapt fastest.

Managing business security in the age of AI

CISO accountability in 2026 has expanded far beyond controls and tooling. Whether we asked for it or not, we now own outcomes tied to business resilience, AI trust, cloud assurance, and continuous availability. The role is less about certainty and more about recovering control in an environment that keeps accelerating.

Every major 2026 initiative – AI agents, third-party risk, cloud, or comms protection – connects to a single board-level question: Are we still in control as complexity and automation scale faster than humans?

Attackers are not just getting more sophisticated; they are becoming more automated. AI changes the economics of attack, lowering cost and increasing speed. That asymmetry is what CISOs are being measured against.

CISOs are no longer evaluated on tool coverage, but on the ability to assure outcomes – trust in AI adoption, resilience across cloud and identity, and being able to respond to unknown and unforeseen threats.

Boards are now explicitly asking whether we can defend against AI-driven threats. No one can predict every new behavior – survival depends on detecting malicious deviations from normal fast and responding autonomously.

Agents introduce decision-making at machine speed. Governance, CI/CD scanning, posture management, red teaming, and runtime detection are no longer differentiators but the baseline.

Cloud security is no longer architectural, it is operational. Identity, control planes, and SaaS exposure now sit firmly with the CISO.

AI-speed threats already reshaping security in 2026

We’re already seeing clear examples of how quickly the threat landscape has shifted in 2026. Darktrace’s work on React2Shell exposed just how unforgiving the new tempo is: a honeypot stood up with an exposed React was hit in under two minutes. There was no recon phase, no gradual probing – just immediate, automated exploitation the moment the code appeared publicly. Exposure now equals compromise unless defenses can detect, interpret, and act at machine speed. Traditional operational rhythms simply don’t map to this reality.

We’re also facing the first wave of AI-authored malware, where LLMs generate code that mutates on demand. This removes the historic friction from the attacker side: no skill barrier, no time cost, no limit on iteration. Malware families can regenerate themselves, shift structure, and evade static controls without a human operator behind the keyboard. This forces CISOs to treat adversarial automation as a core operational risk and ensure that autonomous systems inside the business remain predictable under pressure.

The CVE-2026-1731 BeyondTrust exploitation wave reinforced the same pattern. The gap between disclosure and active, global exploitation compressed into hours. Automated scanning, automated payload deployment, coordinated exploitation campaigns, all spinning up faster than most organizations can push an emergency patch through change control. The vulnerability-to-exploit window has effectively collapsed, making runtime visibility, anomaly detection, and autonomous containment far more consequential than patching speed alone.

These cases aren’t edge scenarios; they represent the emerging norm. Complexity and automation have outpaced human-scale processes, and attackers are weaponizing that asymmetry.

The real differentiator for CISOs in 2026 is less about knowing everything and more about knowing immediately when something shifts – and having systems that can respond at the same speed.

[related-resource]

About the author

Mike Beck

Global CISO

Blog

Network

February 19, 2026

CVE-2026-1731: How Darktrace Sees the BeyondTrust Exploitation Wave Unfolding

Note: Darktrace's Threat Research team is publishing now to help defenders. We will continue updating this blog as our investigations unfold.

Background

On February 6, 2026, the Identity & Access Management solution BeyondTrust announced patches for a vulnerability, CVE-2026-1731, which enables unauthenticated remote code execution using specially crafted requests. This vulnerability affects BeyondTrust Remote Support (RS) and particular older versions of Privileged Remote Access (PRA) [1].

A Proof of Concept (PoC) exploit for this vulnerability was released publicly on February 10, and open-source intelligence (OSINT) reported exploitation attempts within 24 hours [2].

Previous intrusions against Beyond Trust technology have been cited as being affiliated with nation-state attacks, including a 2024 breach targeting the U.S. Treasury Department. This incident led to subsequent emergency directives from the Cybersecurity and Infrastructure Security Agency (CISA) and later showed attackers had chained previously unknown vulnerabilities to achieve their goals [3].

Additionally, there appears to be infrastructure overlap with React2Shell mass exploitation previously observed by Darktrace, with command-and-control (C2) domain avg.domaininfo[.]top seen in potential post-exploitation activity for BeyondTrust, as well as in a React2Shell exploitation case involving possible EtherRAT deployment.

Darktrace Detections

Darktrace’s Threat Research team has identified highly anomalous activity across several customers that may relate to exploitation of BeyondTrust since February 10, 2026. Observed activities include:

Outbound connections and DNS requests for endpoints associated with Out-of-Band Application Security Testing; these services are commonly abused by threat actors for exploit validation. Associated Darktrace models include:

Compromise / Possible Tunnelling to Bin Services

Suspicious executable file downloads. Associated Darktrace models include:‍

Anomalous File / EXE from Rare External Location

Outbound beaconing to rare domains. Associated Darktrace models include:

Compromise / Agent Beacon (Medium Period)
Compromise / Agent Beacon (Long Period)
Compromise / Sustained TCP Beaconing Activity To Rare Endpoint
Compromise / Beacon to Young Endpoint
Anomalous Server Activity / Rare External from Server
Compromise / SSL Beaconing to Rare Destination

Unusual cryptocurrency mining activity. Associated Darktrace models include:

Compromise / Monero Mining
Compromise / High Priority Crypto Currency Mining

And model alerts for:

Compromise / Rare Domain Pointing to Internal IP

IT Defenders: As part of best practices, we highly recommend employing an automated containment solution in your environment. For Darktrace customers, please ensure that Autonomous Response is configured correctly. More guidance regarding this activity and suggested actions can be found in the Darktrace Customer Portal.

Appendices

Potential indicators of post-exploitation behavior:

· 217.76.57[.]78 – IP address - Likely C2 server

· hXXp://217.76.57[.]78:8009/index.js - URL - Likely payload

· b6a15e1f2f3e1f651a5ad4a18ce39d411d385ac7 - SHA1 - Likely payload

· 195.154.119[.]194 – IP address – Likely C2 server

· hXXp://195.154.119[.]194/index.js - URL – Likely payload

· avg.domaininfo[.]top – Hostname – Likely C2 server

· 104.234.174[.]5 – IP address - Possible C2 server

· 35da45aeca4701764eb49185b11ef23432f7162a – SHA1 – Possible payload

· hXXp://134.122.13[.]34:8979/c - URL – Possible payload

· 134.122.13[.]34 – IP address – Possible C2 server

· 28df16894a6732919c650cc5a3de94e434a81d80 - SHA1 - Possible payload

References:

1. https://nvd.nist.gov/vuln/detail/CVE-2026-1731

2. https://www.securityweek.com/beyondtrust-vulnerability-targeted-by-hackers-within-24-hours-of-poc-release/

3. https://www.rapid7.com/blog/post/etr-cve-2026-1731-critical-unauthenticated-remote-code-execution-rce-beyondtrust-remote-support-rs-privileged-remote-access-pra/

About the author

Emma Foulger

Global Threat Research Operations Lead

Your data. Our AI.

Elevate your network security with Darktrace AI

Get a demo

Check out this article by Darktrace: Exploring AI Threats: Package Hallucination Attacks

Exploring AI Threats: Package Hallucination Attacks

AI tools open doors for threat actors

What is an AI hallucination?

Why are AI hallucinations a problem?

Use of generative AI in software development

Malicious packages as attack vectors

How could Darktrace detect an AI Package Hallucination Attack?

Conclusion

References

CVE-2026-1731: How Darktrace Sees the BeyondTrust Exploitation Wave Unfolding

What the Darktrace Annual Threat Report 2026 Means for Security Leaders

Enjoying the blog?

More in this series

What the Darktrace Annual Threat Report 2026 Means for Security Leaders

How AI is redefining cybersecurity and the role of today’s CIO

The State of AI Cybersecurity 2026: Unveiling insights from over 1,500 security leaders

Blog

AI

February 26, 2026

What the Darktrace Annual Threat Report 2026 Means for Security Leaders

The challenge for today’s CISOs

Strategic takeaways from the Annual Threat Report

Managing business security in the age of AI

AI-speed threats already reshaping security in 2026

Blog

Network

February 19, 2026

CVE-2026-1731: How Darktrace Sees the BeyondTrust Exploitation Wave Unfolding

Background

Darktrace Detections

Appendices

Potential indicators of post-exploitation behavior:

References: