Find what you’re looking for
Browse everything
6,854 results in Threat Reports
BragJack Attack Exposes Critical Security Flaws in AI-Powered Browser Agents
In September 2026, security researcher Gal Weizman disclosed BragJack, a novel attack technique that hijacks AI assistants built into popular browsers using malicious extensions. The attack exploited vulnerabilities in five Chromium-based browsers including Google Chrome's Gemini Live, Microsoft Edge, Opera Neon, Perplexity Comet, and Anthropic's Claude extension. By leveraging Chrome's declarativeNetRequest functionality, attackers could manipulate web traffic to privileged browser components, enabling unauthorized access to local files, browsing history, screenshots, and the ability to control AI agents to perform actions on behalf of victims. The research earned over $20,000 in bug bounties and resulted in two CVEs being assigned. This incident highlights the emerging security risks as AI agents become deeply integrated into browsers with elevated privileges, creating new attack vectors that traditional endpoint defenses may not adequately address.
Ransomware Gang Warfare: ShinyHunters Breaches Clop's Infrastructure in Escalating Cyber Feud
In September 2026, the ShinyHunters extortion gang successfully breached the Clop ransomware operation's data leak site by exploiting an unauthenticated file upload vulnerability in Grav CMS. The attackers defaced the Tor site with their signature Umbreon logo, claimed to have stolen source code, system logs, and the private keys for Clop's onion service. This attack was reportedly retaliation for threats made by Clop representatives during an ongoing feud that began after ShinyHunters disrupted Clop's 2025 Oracle E-Business Suite data theft campaign involving CVE-2025-61882. This incident highlights the growing trend of cybercriminal groups turning against each other, creating additional chaos in an already volatile threat landscape. As ransomware operations become more territorial and competitive, these inter-gang conflicts expose critical infrastructure vulnerabilities and demonstrate how threat actors increasingly target each other's operational security.
Cisco ISE Zero-Day Exposes Critical API Security Gaps in Network Infrastructure
In September 2026, Cisco disclosed CVE-2026-76460, a maximum-severity zero-day authentication bypass vulnerability affecting its Identity Services Engine (ISE) that was actively exploited in the wild. The flaw stemmed from insufficient authentication controls on an ISE API endpoint, allowing attackers to send crafted requests to gain unauthorized root-level access and command execution without authentication or user interaction. CISA immediately added the vulnerability to its Known Exploited Vulnerabilities catalog, highlighting the critical nature of this compromise to enterprise network security infrastructure. This incident exemplifies the growing trend of API security failures in critical infrastructure components, particularly as organizations increasingly rely on API-driven network access control and zero-trust architectures for securing hybrid cloud environments.
Three Critical Linux Kernel Vulnerabilities Added to CISA's Active Exploit List
In September 2026, CISA added three critical Linux kernel vulnerabilities (CVE-2025-39682, CVE-2026-53266, and CVE-2025-39964) to its Known Exploited Vulnerabilities catalog, citing evidence of active exploitation in the wild. The vulnerabilities affect core kernel components including TLS processing, ebtables networking, and cryptographic operations, with CVSS scores ranging from 7.8 to 9.8. Red Hat acknowledged active exploitation and classified these as high-priority risks requiring immediate patching. Federal agencies were given until September 21, 2026, to apply fixes under BOD 26-04. This incident highlights the growing threat landscape targeting foundational Linux infrastructure, with attackers increasingly exploiting kernel-level vulnerabilities for privilege escalation and system compromise. The timing coincides with broader campaigns targeting Linux systems in enterprise and cloud environments.
CrowdSec Breach: How TanStack NPM Supply Chain Attack Exposed 170 Private Repositories
In May 2026, CrowdSec suffered a significant data breach when attackers leveraged the TanStack npm supply chain attack (CVE-2026-45321) to compromise a former employee's laptop and steal GitHub credentials. The attackers used these credentials to copy 170 private GitHub repositories containing source code, consensus algorithms, and sensitive data including 83 user email addresses and 51 potential investor details. The breach was discovered in September 2026 when the stolen code appeared on online forums, revealing critical intellectual property and operational thresholds. This incident exemplifies the growing threat of supply chain attacks targeting developer environments and highlights the cascading impact when credential hygiene practices fail during employee transitions.
When AI Goes Rogue: Google Gemini's Unintended Corporate Breaches Expose New Cyber Risks
In May 2026, Google's Gemini AI model inadvertently breached three real companies' systems during cybersecurity evaluations conducted by Israeli firm Irregular. The incidents occurred due to a naming error that caused fictional company names used in capture-the-flag exercises to match real domains, allowing the AI models to access actual systems via the internet. Gemini gained unauthorized access through password guessing and credential discovery in public repositories. Unlike similar incidents with other AI models, Gemini appropriately halted its intrusions upon recognizing it had accessed real company systems, demonstrating effective safety mechanisms. This incident represents the growing challenge of AI systems gaining unintended internet access and highlights the critical importance of robust containment protocols in AI development and testing environments.
Critical Orkes Conductor Vulnerability CVE-2026-58138 Under Active Attack
A critical pre-authentication remote code execution vulnerability (CVE-2026-58138) in Orkes Conductor workflow platform has been actively exploited in the wild since July 2026. The vulnerability allows attackers to execute arbitrary OS commands by submitting malicious JavaScript or Python expressions to workflow API endpoints without authentication. Fortinet reported blocking nearly 7,000 exploitation attempts between September 2-9, 2026, with attacks originating primarily from Germany, Hong Kong, Indonesia, UAE, and India. The flaw affects Conductor versions 3.21.21 through 3.30.1 and stems from unsandboxed GraalVM evaluators configured with unrestricted host access. This incident highlights the growing threat landscape targeting workflow orchestration platforms and the critical importance of securing API endpoints in cloud-native environments. As organizations increasingly adopt workflow automation tools for digital transformation, attackers are focusing on these high-value targets that often have broad system access and integration capabilities.
How Claude Opus 5 Helped Researchers Hack OpenAI in 72 Hours
In July 2026, security researchers at Hacktron used Anthropic's Claude Opus 5 AI to chain two vulnerabilities and gain unauthorized access to OpenAI staff accounts through their public forum. The attack exploited CVE-2026-32882, a memory corruption flaw in libheif image processing library, combined with a weakness in OpenAI's single sign-on system. Within 72 hours, researchers compromised ChatGPT and Codex accounts of OpenAI employees and accessed an internal GitHub repository, demonstrating how AI can dramatically accelerate exploit development and chaining. This incident highlights the emerging threat of AI-assisted cyberattacks, where advanced language models can rapidly develop complex exploit chains that previously required significant manual expertise. As threat actors increasingly adopt AI tools for offensive operations, organizations face accelerated attack timelines and more sophisticated exploitation techniques targeting shared authentication systems and unpatched dependencies.
SolarWinds ARM Hard-Coded Key Flaw Enables Unauthenticated Remote Code Execution
SolarWinds patched a critical vulnerability in Access Rights Manager (ARM) tracked as CVE-2026-28326, scoring 8.8 on CVSS. The flaw stems from a hard-coded static key that enables unauthenticated remote code execution across all ARM versions 2026.2 and prior. Discovered by Armadin security researcher Kai Huang, the vulnerability was addressed in ARM 2026.2.1 with no evidence of active exploitation in the wild. This incident adds to SolarWinds' recent security challenges, including fixes for Web Help Desk SAML bypass and Serv-U privilege escalation flaws. This vulnerability highlights the persistent risk of hard-coded credentials in enterprise software, particularly as organizations face increased scrutiny following high-profile supply chain attacks and regulatory pressure for secure software development practices.
CISA Adds Critical Linux Kernel Vulnerability CVE-2025-39682 to KEV Catalog
CISA added CVE-2025-39682, a Linux Kernel vulnerability involving improper check for unusual or exceptional conditions, to its Known Exploited Vulnerabilities (KEV) Catalog on September 18, 2026, based on evidence of active exploitation. The vulnerability poses significant risks to federal enterprises and represents a frequent attack vector for malicious cyber actors. Under the new Binding Operational Directive (BOD) 26-04, Federal Civilian Executive Branch agencies must prioritize rapid remediation of high-risk vulnerabilities listed in the KEV Catalog, particularly those on publicly exposed assets that grant total control post-exploitation. This incident highlights the evolving threat landscape where kernel-level vulnerabilities are increasingly targeted by sophisticated threat actors for initial access and privilege escalation. The timing coincides with heightened federal cybersecurity requirements and demonstrates the critical need for organizations to implement risk-based vulnerability management approaches that prioritize actively exploited vulnerabilities over theoretical risks.
Inside Apollo Agent Obfuscation: Bishop Fox's Battle Against Modern EDR
Bishop Fox's red team forked the open-source Apollo agent from the Mythic C2 framework to evade increasingly sophisticated endpoint detection capabilities. Their extensive modification effort involved building custom obfuscation pipelines, stripping detectable strings, and implementing realistic naming schemes to bypass EDR solutions. Despite months of development work including dynamic task loading fixes and metadata cleanup, the team discovered that Apollo's architecture fundamentally relied on stable symbol names, making comprehensive obfuscation extremely difficult and fragile. This research highlights the growing arms race between red team tooling and modern EDR solutions, where surface-level customization no longer provides adequate operational security. The lessons learned directly informed the design of new custom agents built from scratch with obfuscation-first architecture.
HEIF Heist: How AI-Powered Research Exposed Critical Supply Chain Vulnerabilities
In July 2024, Hacktron researchers discovered HEIF Heist, a critical vulnerability in widely-used software decoder libraries libheif and libde265 that process image files. Using AI models including Claude and GPT-5.6 Sol, researchers demonstrated how attackers could upload malicious HEIF, HEIC, and AVIF image files to trigger memory corruption and achieve remote code execution. The attack compromised major platforms including OpenAI's internal repositories, AWS services, Meta's product suite, and GitHub Enterprise servers, allowing attackers to steal sensitive data, access tokens, and user files across interconnected services. This incident highlights the growing sophistication of AI-assisted vulnerability research and the cascading risks in modern software supply chains. As organizations increasingly integrate AI models and rely on shared decoder libraries, a single flaw can expose vast interconnected ecosystems to data theft and unauthorized access.