GitHub Copilot’s new coding agents, which are integrated into IDEs, are susceptible to a specific type of “workflow-level” jailbreak attacks. These attacks can bypass chat refusals, allowing agents to generate harmful code while performing standard software development tasks unwittingly. According…
Category: GBHackers Security | #1 Globally Trusted Cyber Security News Platform
HalluSquatting Attack Lets Hackers Turn AI Coding Assistants Into Botnet Installers
A newly disclosed attack technique called “HalluSquatting” is raising serious concerns in the AI security landscape. This technique demonstrates how attackers can exploit large language model (LLM) hallucinations to covertly compromise systems and potentially create botnets on a large scale.…
npm and PyPI Malware Campaign Exfiltrates CI/CD Secrets Through Fake Payment SDKs
A coordinated supply-chain campaign that pushed 17 malicious packages across npm and PyPI, masquerading as SDKs for well-known payment services including PaySafe, Skrill and Neteller. The campaign’s packages 17 npm modules published with four rapid versions each and four PyPI…
SNOW Malware Ecosystem Uses Teams Phishing, WebSocket Tunnels, and Browser Extensions
Threat actors are increasingly chaining classic phishing with collaboration platforms and covert tunneling to create highly believable intrusion paths. A recent multi-stage campaign attributed to UNC6692 exposes how adversaries combine email bombardment, Microsoft Teams impersonation, malicious browser extensions, WebSocket tunnels,…
Everest Ransomware Encryptor Uses ConfuserEx-Protected .NET Binary With Wake-on-LAN Capability
A recent technical analysis of an Everest ransomware encryptor reveals a purpose-built, ConfuserEx-protected .NET 4.0 binary that combines heavy obfuscation, misleading cryptographic declarations, and uncommon network tactics to maximize impact and impede response. The analyzed sample (hlntqyun.exe, SHA-256 1df92b…) is…
Umbrij Malware Lets ToddyCat Hackers Hijack Gmail Accounts Through Google API Abuse
A targeted campaign in which the ToddyCat (aka APT-style) group leverages a previously observed loader family, Umbrij, to hijack Gmail accounts by abusing Google APIs. Chaining that capability to broad remote access achieved through a malicious MSI installer masquerading as…
New Helix Extortion Group Targets Enterprises With MFA Abuse and SharePoint Exfiltration
A previously unreported data extortion operation dubbed “Helix” that targets enterprises using identity-focused entry techniques and automated SharePoint exfiltration. The group’s playbook combines voice phishing (vishing), device-code phishing to capture session tokens and bypass Conditional Access controls, rapid MFA registration…
Claude AI Prompt Injection Attack Turns Chatbot Into Stealthy C2 Agent to Achieve Remote Code Execution
Claude Desktop’s synced Personal Preferences feature can be exploited as a covert prompt-injection vector, transforming the AI assistant into a de facto command-and-control (C2) agent. This method allows for remote code execution on a compromised user workstation without the need…
DuckDuckGo Browser Blocks YouTube Ads Using uBlock Origin Filter Lists
DuckDuckGo has quietly expanded its privacy-first browser capabilities by introducing a YouTube ad-blocking feature. This feature uses community-driven uBlock Origin filter lists to detect and remove video ads. From a security and privacy standpoint, this approach is significant because it…
Best Next-Generation Firewall (NGFW) Solutions Compared (2026): Features & Pricing
Twelve firewalls, one question: which NGFW earns a place at your network edge in 2026? For most enterprises the shortlist starts with Fortinet FortiGate (best price-performance) and Palo Alto Networks (deepest application control), but the right answer shifts with your size, region,…
Claude Cowork Expands to Web and Mobile With Background AI Agent Workflows
Claude Cowork was expanded beyond the desktop, rolling out web and mobile versions that let AI-driven task sessions persist across devices and continue running in the background without an active connection. The beta rollout begins with Max-tier subscribers over the…
Attackers Can Generate Duplicate Verified GitHub Commits Using Signature Malleability
Attackers can silently clone “Verified” GitHub commits by abusing signature malleability in Git’s commit-signing formats, creating byte‑different commits with identical content, valid signatures, and fresh “Verified” badges under new hashes. This breaks the long‑standing assumption that a verified commit hash…
Fancy Bear Uses LSB Steganography and Reflective Loading to Run C# Remote-Control Trojan
A new intrusion campaign attributed to APT‑C‑20 (aka Fancy Bear, APT28) demonstrates the group’s continued refinement of stealthy, fileless techniques: weaponized Office documents that deploy a COM‑hijacking DLL. Extract shellcode hidden via LSB steganography in a PNG, and use reflective…
IonStack Exploit Chain Lets Hackers Root Android 17 Phones With a Single URL Click
Nebula Security has revealed a significant exploit chain known as “IonStack,” demonstrating how attackers could gain full root access on Android 17 devices with just a single click on a malicious URL. This raises serious concerns about browser-to-kernel attack surfaces…
ESET Threat Report H1 2026 Highlights Malicious AI Skills, ClickFix Surge, and EDR Killers
ESET’s H1 2026 threat report shows attackers accelerating the use of familiar playbooks with AI-flavored lures, social engineering, and defense evasion rather than inventing entirely new ones. The clearest signals are the rise of malicious AI skills, a doubled ClickFix…
Exposed Banana RAT Infrastructure Reveals Payload Generator and Obfuscator Tooling
A publicly indexed server at 198[.]245[.]53[.]26, discovered via Shodan, exposed more than simple staging files it revealed an active payload-generation backend and obfuscation tooling tied to two distinct Banana RAT branches. The host served static stages (st.txt, payload.php) and a…
CrowdStrike Uncovers 5 New Prompt Injection Techniques Targeting AI Agents
CrowdStrike has identified five new techniques for prompt injection targeting AI agents, emphasizing the rapid evolution of adversarial methods as enterprises increasingly deploy autonomous AI systems. Detailed in a report published on July 7, 2026, by CrowdStrike’s AI security research…
Discord Confirms Bug That Incorrectly Banned 8,200 Users Since May 2026
Discord has confirmed a significant flaw in its automated moderation and enforcement pipeline that led to the wrongful banning of approximately 8,200 user accounts between May and early July 2026. This raises concerns about the reliability of AI-assisted trust and…
AI-as-a-Service Botnet Routes Malicious Workloads Across Compromised Windows and Linux Hosts
The underground advertisement for the so-called Mycelium Framework reads like another feature‑packed botnet sales pitch: cross‑platform payloads, encrypted C2, persistence, exploit modules, credential theft, and lateral movement. Those building blocks are not new. What makes Mycelium notable is its advertised…
Claude Code, Cursor, and OpenAI Codex Trigger Cyberattack-Like Telemetry Alerts
AI-powered coding assistants such as Claude Code, Cursor, and OpenAI Codex are increasingly triggering endpoint detection and response (EDR) alerts that resemble active cyberattacks, according to new research from Sophos X-Ops. This analysis, based on real-world telemetry collected in June…