✨ No need to do from scratch. Deploy a Validated Containment Architecture built for your AI platform. →Deploy a Validated Containment Architecture for your AI platform. →A Validated Containment Architecture for your AI platform. →Validated Containment Architectures are here. →Contain Threats NowExplore✨
Computer Software/Engineering
Breach intelligence, attack campaigns, and threat reports targeting the Computer Software/Engineering sector.
Explore Other Sectors
Computer Software/Engineering Threat Reports
Critical LLM Vulnerability: System Prompt Extraction via Form Fields
In January 2026, security researchers identified a critical vulnerability in Large Language Models (LLMs) integrated into AI assistants. Despite implementing architectural constraints to restrict chatbots to templated responses, attackers exploited the models' ability to populate form fields, enabling the extraction of system prompts. This method allowed unauthorized access to sensitive information, bypassing traditional output restrictions. The incident underscores the evolving nature of prompt injection attacks and the necessity for comprehensive security measures in AI deployments. As AI integration becomes more prevalent, understanding and mitigating such vulnerabilities is crucial to maintaining data integrity and user trust.
6 months ago
Kill Chain
Meta AI's 2024 Chatbot Vulnerability Exposes User Data
In December 2024, a critical vulnerability was discovered in Meta AI's chatbot platform, allowing unauthorized users to access private prompts and AI-generated responses of other users. The flaw involved manipulating unique identifiers in the browser's network traffic, enabling access to others' chatbot interactions without proper permission checks. Meta was notified on December 26, 2024, and patched the vulnerability by January 24, 2025. There was no evidence of the flaw being exploited in the wild. This incident underscores the importance of securing AI chatbot platforms against unauthorized access, especially as they handle sensitive user data. Organizations must implement robust authorization mechanisms and regularly audit their systems to prevent similar vulnerabilities.
6 months ago
Kill Chain
Introducing Julius: Enhancing AI Security with LLM Service Fingerprinting
In January 2026, Praetorian released Julius, an open-source tool designed to identify and fingerprint Large Language Model (LLM) services across corporate networks. Julius enables security teams to detect various LLM services, such as Ollama, LiteLLM, and Open WebUI, by analyzing endpoints and extracting information about the models in use. This tool addresses the growing challenge of unmanaged and potentially insecure LLM deployments within organizations, which can be exploited by attackers for unauthorized access, data exfiltration, or lateral movement within networks. The release of Julius is particularly timely given the increasing integration of LLMs into enterprise environments and the associated security risks. Recent incidents have highlighted vulnerabilities in LLM applications, including prompt injection attacks and data leakage, underscoring the need for robust detection and monitoring tools like Julius to enhance organizational security postures.
6 months ago
Kill Chain
ServiceNow's 'BodySnatcher' Vulnerability: A Wake-Up Call for AI Security
In October 2025, ServiceNow identified and patched a critical vulnerability, CVE-2025-12420, in its AI Platform. Dubbed 'BodySnatcher,' this flaw allowed unauthenticated attackers to impersonate legitimate users, potentially granting unauthorized access to sensitive enterprise data. The vulnerability affected components such as Now Assist AI Agents and the Virtual Agent API, with ServiceNow deploying security updates to hosted instances and providing patches for self-hosted customers. ([techradar.com](https://www.techradar.com/pro/security/servicenow-patches-critical-security-flaw-which-could-allow-user-impersonation?utm_source=openai)) This incident underscores the growing security challenges associated with AI-driven platforms. Organizations are urged to promptly apply security updates and review their AI system configurations to mitigate similar risks. ([techradar.com](https://www.techradar.com/pro/security/servicenow-patches-critical-security-flaw-which-could-allow-user-impersonation?utm_source=openai))
6 months ago
Kill Chain
Google's 2026 Disruption of IPIDEA Proxy Network
In January 2026, Google’s Threat Intelligence Group (GTIG) disrupted IPIDEA, a China-based residential proxy network that covertly enrolled millions of consumer devices into its infrastructure. By embedding malicious software development kits (SDKs) into various applications, IPIDEA transformed devices into proxy nodes without user consent, facilitating cybercriminal activities such as password spraying and unauthorized access to cloud environments. Google’s intervention, which included legal actions to seize control domains and collaboration with partners like Cloudflare and Lumen’s Black Lotus Labs, resulted in a significant reduction of IPIDEA’s operational capacity, removing millions of devices from the network. ([blog.google](https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/gtig-ipidea-disrupted/?utm_source=openai)) This incident underscores the escalating misuse of residential proxy networks by cybercriminals and state-sponsored actors. The proliferation of such networks highlights the urgent need for enhanced vigilance among developers and consumers regarding the integration and use of third-party SDKs, as well as the importance of industry-wide collaboration to dismantle malicious infrastructures.
6 months ago
Kill Chain
Google Engineer Convicted of AI Trade Secrets Theft for China Startup
In January 2026, former Google engineer Linwei Ding was convicted on seven counts of economic espionage and seven counts of theft of trade secrets. Between May 2022 and April 2023, Ding illicitly transferred over 2,000 confidential documents related to Google's AI technology to his personal Google Cloud account. These documents detailed proprietary information about Google's supercomputing data center infrastructure, including custom Tensor Processing Unit chips, Graphics Processing Unit systems, and the Cluster Management System software. During this period, Ding secretly affiliated with two China-based technology companies, including founding Shanghai Zhisuan Technologies Co., while still employed at Google. He employed deceptive tactics to conceal his activities, such as copying data into the Apple Notes application and converting them to PDFs before uploading them to his personal account. The scheme was uncovered when Google discovered Ding's public presentation in China to potential investors about his startup. This case underscores the persistent threat of insider threats and economic espionage, particularly in the competitive field of artificial intelligence. Organizations must remain vigilant in protecting their intellectual property and sensitive information from both internal and external threats. The incident highlights the importance of robust security measures and monitoring systems to detect and prevent unauthorized access and data exfiltration.
6 months ago
Kill Chain
Malicious Chrome Extensions Compromise User Security by Hijacking Affiliate Links and Stealing ChatGPT Tokens
In January 2026, cybersecurity researchers uncovered a series of malicious Google Chrome extensions designed to hijack affiliate links and steal OpenAI ChatGPT authentication tokens. Notably, the 'Amazon Ads Blocker' extension, uploaded by '10Xprofit' on January 19, 2026, claimed to block Amazon ads but covertly injected the developer's affiliate tag into product links, replacing existing ones. This extension was part of a larger cluster targeting e-commerce platforms like AliExpress, Amazon, Best Buy, Shein, Shopify, and Walmart. Additionally, 16 other extensions masquerading as ChatGPT productivity tools were found to exfiltrate ChatGPT session tokens, granting attackers full access to users' conversation histories and associated data. This incident underscores the growing trend of malicious browser extensions exploiting the popularity of AI tools and e-commerce platforms. The deceptive nature of these extensions, often appearing legitimate and even bearing 'Featured' badges, highlights the need for heightened vigilance among users and stricter vetting processes by browser extension stores to prevent such security breaches.
6 months ago
Kill Chain
OpenClaw AI's Security Challenges in 2026: A Wake-Up Call for AI Deployment
In January 2026, the open-source AI assistant OpenClaw (formerly known as Clawdbot and Moltbot) faced significant security vulnerabilities due to its deep system access and widespread misconfigurations. Users inadvertently exposed control panels and API keys, leading to unauthorized access and potential data breaches. Additionally, the platform's susceptibility to prompt injection attacks allowed malicious actors to manipulate the AI into executing unintended commands, posing severe risks to user data and system integrity. ([axios.com](https://www.axios.com/2026/01/29/moltbot-cybersecurity-ai-agent-risks?utm_source=openai)) The incident underscores the critical need for robust security measures in AI deployments, especially as autonomous agents gain popularity. Organizations must prioritize secure configurations, implement strict access controls, and conduct regular security audits to mitigate risks associated with AI assistants like OpenClaw.
6 months ago
Kill Chain
Critical Unauthenticated RCE Vulnerability in n8n (CVE-2026-21858)
In early January 2026, a critical vulnerability, CVE-2026-21858, was discovered in n8n, an open-source workflow automation platform. This flaw, dubbed 'Ni8mare,' allows unauthenticated remote code execution due to improper input validation in form-based workflows. Exploiting this vulnerability enables attackers to gain full control over affected servers, leading to potential data theft, workflow manipulation, and system compromise. ([purple-ops.io](https://www.purple-ops.io/resources-hottest-cves/cve-2026-21858-ai-exploit/?utm_source=openai)) The widespread use of n8n, with over 100 million Docker pulls and numerous internet-exposed instances, amplifies the risk. As of January 11, 2026, approximately 59,500 vulnerable hosts remain exposed, highlighting the urgency for organizations to apply the available patches promptly. ([techradar.com](https://www.techradar.com/pro/security/thousands-of-n8n-instances-under-threat-from-top-security-issue?utm_source=openai))
6 months ago
Kill Chain
Safeguarding AI Assets: Lessons from the 2025 Model Extraction Attack
In 2025, a significant AI model extraction attack was identified, where adversaries systematically queried a proprietary machine learning model's API to replicate its functionality. By sending carefully crafted inputs and analyzing the outputs, attackers reconstructed a substitute model that closely mirrored the original's behavior. This breach exposed the model's intellectual property, leading to potential competitive disadvantages and financial losses for the organization. The incident underscores the vulnerabilities inherent in exposing AI models through APIs without adequate security measures. ([techtarget.com](https://www.techtarget.com/searchsecurity/tip/AI-model-theft-Risk-and-mitigation-in-the-digital-era?utm_source=openai)) The rise of such model extraction attacks highlights the urgent need for organizations to implement robust defenses, including rate limiting, output perturbation, and behavioral monitoring, to protect their AI assets from unauthorized replication and misuse. ([snyk.io](https://snyk.io/articles/ai-model-theft/?utm_source=openai))
6 months ago
Kill Chain
Swarmer Tool: Exploiting Windows Legacy Features for Stealthy Registry Persistence
In February 2025, Praetorian Inc. introduced 'Swarmer,' a tool designed to achieve stealthy Windows registry persistence without triggering Endpoint Detection and Response (EDR) systems. By exploiting legacy Windows features such as mandatory user profiles and the Offline Registry API, Swarmer allows low-privilege users to modify the NTUSER hive covertly. This method bypasses standard registry APIs monitored by EDR solutions, enabling attackers to establish persistence without detection. The release of Swarmer underscores the ongoing challenges in cybersecurity, particularly the exploitation of overlooked system functionalities. As attackers continue to innovate, it is imperative for organizations to reassess and fortify their security postures against such sophisticated techniques.
6 months ago
Kill Chain
Match Group's 2026 Data Breach: A Wake-Up Call for Digital Security
In late January 2026, Match Group, the parent company of popular dating platforms such as Hinge, Match.com, and OkCupid, experienced a significant data breach orchestrated by the cybercriminal group ShinyHunters. The attackers claimed to have exfiltrated over 10 million user records, including user IDs, transaction details, IP addresses, and internal corporate documents. The breach was reportedly facilitated through a vulnerability in AppsFlyer, a mobile marketing analytics platform utilized by Match Group. Match Group promptly initiated an investigation with external cybersecurity experts and began notifying affected users. Preliminary findings indicated that user login credentials, financial information, and private communications were not accessed. ([cybernews.com](https://cybernews.com/security/hinge-okcupid-data-leak-shinyhunters-claims/?utm_source=openai)) This incident underscores the persistent threat posed by sophisticated cybercriminal organizations like ShinyHunters, known for targeting high-profile companies and leaking sensitive data. The breach highlights the critical importance of securing third-party integrations and the need for robust cybersecurity measures to protect user data. Organizations must remain vigilant and proactive in identifying and mitigating potential vulnerabilities to prevent similar incidents.
6 months ago
Kill Chain
Stop Active Cloud Data Exfiltration
Aviatrix Breach Lock helps teams instantly identify what data is leaving the environment, from which workload, and where it’s going — during an active breach.
Looking for threats in a different sector?
Browse All Threat Reports