A vulnerability in ChatGPT's Lockdown Mode allows attackers to bypass its protections and perform silent exfiltration of chat data, phishing, and malware delivery. The flaw exploits the HTML page preview feature in ChatGPT code blocks, enabling attackers to open pop-up windows outside the sandbox, exfiltrate entire chat histories without user interaction, and display phishing content. This bypass undermines Lockdown Mode's restrictions on outbound network requests and content loading. The attack surface includes data theft, social engineering via interface manipulation, and malware distribution within the trusted ChatGPT interface.
Critical9.2KEVEPSS 82%Check Point ResearchCVE-2026-111497
This report summarizes multiple significant cyber threats and incidents disclosed in early November 2026. Major breaches include a $6.3 billion cryptocurrency theft from Binance via compromised internal infrastructure, a malicious JavaScript attack on Polymarket's platform, and data breaches affecting Japan Airlines and the Florida Department of Education. The report also highlights the use of AI by threat actors such as the GOLD OGRE ransomware group and North Korean cyber operators to enhance attack capabilities. Several critical vulnerabilities are noted, including CVE-2026-111497 (HTTP2Shell) affecting Brotli decompression, a critical Linux kernel TCP networking flaw (CVE-2026-98432), and a Drupal Commerce Core access control vulnerability (CVE-2026-118539). Cisco patched a high-severity SSL VPN crash vulnerability (CVE-2026-76418) that was exploited in the wild. Additional threat intelligence includes malware campaigns and advanced persistent threat activity targeting various sectors globally.
A malicious version of the Claude Code extension (versions 2.1.312 and 2.1.313) was briefly available on the Cursor marketplace, embedding a JavaScript loader that attempts to establish persistent SSH backdoor access. The loader collects host information, adds an attacker-controlled SSH public key to authorized_keys, enables the local SSH service, creates a reverse SSH tunnel, configures persistence mechanisms, and attempts to exfiltrate host details. The incident is limited to the Cursor marketplace; official Microsoft Visual Studio Marketplace and npm packages were not affected. Successful remote access depends on host privileges and SSH configuration. Organizations should identify affected systems, preserve evidence, remove malicious artifacts, and rotate credentials if exposure is suspected.
Cisco Talos has identified a new Windows-based command-and-control (C2) framework named NIK C2 that integrates AI directly into its attack chain. This Go-based implant enables adversaries to dump credentials, capture keystrokes, and deploy screen-locking ransomware. It uses an embedded OpenAI model to analyze harvested data in real time, funding API requests by hijacking the victim's own OpenAI API keys. This integration allows continuous data collection and exploitation without offline delays, turning compromised hosts into active operational assets. The use of victim API keys also masks attacker infrastructure, complicating detection and incident response. Defenders are advised to implement continuous behavioral monitoring and update endpoint defenses accordingly.
Google Project Zero compared two automated vulnerability discovery tools for Chrome's V8 JavaScript engine: Fuzzilli, a traditional fuzzer, and BigSleep, an LLM-based bug finder. Over six months, BigSleep found more vulnerabilities (45) than Fuzzilli (37), with an estimated 25-60% overlap in bugs found. Both tools find unique vulnerabilities, indicating they complement each other rather than one replacing the other. The study recommends using both fuzzing and LLM-based analysis to maximize vulnerability discovery before impacting users.
CVE-2026-111497, dubbed Node2Shell, is a critical out-of-bounds write vulnerability in Google’s Brotli decoder affecting versions 1.0.3 through 1.2.0. It impacts multiple platforms including Node.js, .NET, PowerShell, and Python packages, allowing unauthenticated attackers to cause memory corruption that can lead to process crashes, memory leaks, or arbitrary code execution. The flaw has been patched in Brotli versions 1.0.10, 1.1.1, and 1.2.1, with corresponding updates released by Microsoft, Node.js, and Python package maintainers. No exploitation in the wild has been reported, but proof-of-concept code demonstrates code execution on certain Linux arm64 Node.js builds. The vulnerability arises when attacker-controlled compressed data is processed by the affected Brotli decoder.
Shopware 6.7 contains a stored cross-site scripting (XSS) vulnerability in its media upload functionality that allows attackers with upload access to execute arbitrary JavaScript in an administrator's browser. This can lead to theft of administrator authentication tokens and unauthorized access to the Shopware Administration API. If the compromised administrator has extension management privileges, attackers can upload and activate malicious plugins to achieve remote code execution on the server. No vendor patch or statement is currently available. Mitigation involves restricting media upload access to trusted users or disabling HTML file uploads.
Thelia versions 2.6.0 and 2.6.1 contain multiple vulnerabilities affecting file management, Smarty templating, module installation, and administrative workflows. These include remote code execution, server-side template injection, stored and reflected cross-site scripting, open redirect, and username enumeration. Exploitation requires authenticated administrative access for most issues, but some vulnerabilities enable unauthenticated attackers to enumerate valid admin usernames. No patch is currently available, and risk mitigation involves restricting administrative interface access and disabling or limiting certain features.
This report summarizes multiple cybersecurity incidents and emerging threats including AI-assisted bank breaches in South Korea, a poem-guided botnet called PoeLLM targeting AI and open-source services, a supply chain attack on the Tensorlake npm SDK, and the sentencing of the Empire Market co-founder. The PoeLLM malware uses a novel method of hiding its command-and-control server IP address encoded in a GitHub-hosted poem. The Tensorlake npm SDK was compromised to deliver a credential-stealing worm. Nvidia's DCGM Exporter had a high-severity denial-of-service vulnerability that has been patched. The South Korean bank breaches are under investigation with AI suspected to be involved. The Empire Market co-founder received a 40-year prison sentence for running a dark web marketplace. These events highlight evolving attack techniques and ongoing law enforcement actions.
Since August 2023, a threat actor has conducted a long-running malware campaign on the NPM registry called MALFEX, publishing eight malicious packages that have accumulated over 40,000 downloads. The campaign distributes malware including the Overlord RAT and various infostealers targeting Windows systems. Some malicious packages remain installable as of October 2026, with one package, function-flag, having over 37,000 downloads and no advisory flags. The malware enables remote control, data theft from browsers, Discord clients, and cryptocurrency wallets. The infection routines are designed to fail silently on non-Windows systems, limiting impact to Windows users who directly install these packages. No geographic or organizational targeting has been observed.
Readwise Reader for Android version 8.7.2 contains three stored cross-site scripting (XSS) vulnerabilities. These flaws allow attackers who can supply malicious documents or metadata to execute arbitrary JavaScript within the app's WebView context. The vulnerabilities arise from insufficient sanitization of document metadata fields, improper escaping of URL metadata, and permissive sanitization rules for SVG markup. Exploitation could compromise user data confidentiality and integrity, including access to stored documents, credentials, and session tokens. A patch addressing one of the issues is available in version 8.10.1. Users are advised to update when possible and exercise caution when adding content from untrusted sources.
A malicious NPM package named indexed-btree, impersonating the legitimate sorted-btree package, was discovered to contain hidden malware in its JavaScript prototype method. This supply chain attack accumulated over 2 million weekly downloads by creating a credible GitHub repository and avoiding typical detection methods. The malware collects system information, communicates with a hardcoded Slack channel and Telegram chat, and uses a blockchain smart contract for command-and-control. Multiple related malicious packages have also been identified and removed after millions of downloads. The campaign is ongoing and represents a sophisticated threat to organizations using the NPM ecosystem.
A malware campaign involving the npm package 'indexed-btree' demonstrates how attackers evade traditional supply chain defenses by embedding malicious code in the package's runtime behavior instead of installation scripts. This technique allows the malware to bypass install-time security checks and execute malicious actions during normal package usage.
Brevo experienced a supply-chain attack where attackers stole a Cloudflare API key and used it to inject malicious ClickFix scripts into Brevo's websites and JavaScript files embedded on customer sites. This injection was used to distribute malware to visitors of affected sites.
An investigation revealed a sophisticated cryptocurrency and credential-stealing operation using Amatera stealer as the primary payload. The attack chain leverages Cloudflare Workers to inject JavaScript stored on BNB Smart Chain, presenting fake Google CAPTCHA prompts that lead to WebDAV-based DLL execution. Two parallel infection chains were identified, one loading through a DLL named 'pf.ch' and another through 'verification.google', both delivering Amatera stealer with different secondary payloads. The 'pf.ch' variant deployed ZigCryptoStealer via NativeAOT loader and a Go-based reverse proxy, while the 'verification.google' variant installed NetSupport Manager with C2 infrastructure in Russia. The operation affects numerous countries with primary focus on stealing cryptocurrency wallets, credentials from password managers, and various authentication data through extensive collection rules targeting over 400 applications.
Cisco Talos is tracking a cryptocurrency theft campaign that abuses the Google Visualization API to deliver obfuscated JavaScript from Google Sheets, which victims are socially engineered to inject into their browser sessions. The attack uses a variation of ClickFix social engineering, convincing targets to paste malicious JavaScript into the Chrome address bar or install it via the Tampermonkey extension for persistence. The malicious script acts as a web skimmer, hooking the browser's fetch API to replace cryptocurrency deposit addresses and manipulate clipboard contents, deceiving users with fake bonus interfaces. The campaign targets users on cryptocurrency and hacking forums with a lure about a nonexistent API vulnerability to entice quick financial gain. The actors moved their hosting to Google Docs and Sheets after disruptions on paste sites. While this campaign primarily targets individuals, the techniques used could be adapted for broader supply-chain or web-based attacks.
Improper neutralization of special elements used in an OS command (CWE-78) in the blueprint resynthesis framework in Amazon Web Services codecatalyst-blueprints before 0.3.156 might allow a user with permission to commit to a repository in the project to execute arbitrary commands in the blueprint resynthesis environment via shell metacharacters in the owner field of a [local] merge strategy entry in a crafted .ownership-file.
Version 0.3.156 removes shell interpretation of the owner field, running the command directly rather than through a shell, and rejects values outside an allowlisted command form. This eliminates shell metacharacter command injection. To remediate this issue, users should upgrade to version 0.3.156 or later.
No action is required for use of the Amazon CodeCatalyst service. Resynthesis runs in an isolated per-project environment with scoped credentials, and the service applies server-side validation there that rejects [local] merge strategy commands outside a restricted allowlisted form, including for blueprint versions published before 0.3.156.
JSCeal is a sophisticated cryptocurrency-focused stealer malware delivered as compiled V8 bytecode executed by a bundled Node.js runtime. It employs multiple layers of JavaScript obfuscation and encryption, including RC4 and AES-256-CBC, to protect its payload. The malware's capabilities include keylogging, browser credential theft, cryptocurrency collection, HTTPS traffic interception via a local man-in-the-middle proxy, and session replay using stolen cookies. Recent variants target macOS and newer V8 versions, showing active development. Check Point Research developed a fully static deobfuscation pipeline to analyze JSCeal without executing it, enabling detailed understanding of its behavior. The toolkit is publicly available and supports analysis of multiple samples.
Two men in Australia were arrested for their alleged involvement in TeamPCP, a cybercrime syndicate responsible for extensive software supply chain attacks. TeamPCP embedded malicious code in numerous open source software tools, compromising thousands of global businesses. The group used a self-propagating worm called Shai-Hulud to infect open source projects and steal developer credentials, enabling further malicious code insertion. They also ran a contest to recruit additional hackers by incentivizing supply chain compromises. Their attacks included targeting AI infrastructure via the LiteLLM open source AI gateway, harvesting cloud service keys from over 2,500 organizations. TeamPCP compromised thousands of code repositories on GitHub by exploiting compromised developer tools. The group appears to be a loose coalition of multiple cybercriminal actors rather than a single structured entity.
