AI-powered offensive tools will outpace patch cycles, turning unfixed flaws into fast-moving exploits. The real blockers are business disruption, process delays, and resistance to change. #AIExploit #PatchDelay #RiskOps ift.tt/AH7ZkOp

“Hexstrike‑AI”、LLMと150種以上ツール統合の攻撃オーケストレーションAI、リリース直後にサイバー犯罪者が武器化。Citrixのゼロデイを10分未満でスキャン＆ウェブシェル設置。#HexstrikeAI #AIExploit gbhackers.com/hackers-use-he…

Agentic AI Capture The Flag (CTF) – FinBot DEMO: Goal Manipulation - youtube.com/watch?v=UORcoidb… - github.com/OWASP-ASI/finbot-… by @owasp Explore the OWASP Agentic AI CTF through a hands-on walkthrough of the FinBot demo. This session highlights the “goal manipulation” challenge, revealing how attackers can exploit agentic AI systems and showcasing strategies to identify and defend against these advanced threats. FinBot is part of the OWASP GenAI Security Project’s Agentic Security Initiative, created to equip builders and defenders with hands-on tools for understanding and mitigating agentic AI risks. FinBot is an Agentic Security Capture The Flag (CTF) interactive platform that simulates real-world vulnerabilities in agentic AI systems using a simulated Financial Services-focused application. @e2hln #AgenticAI #AISecurity #OWASP #OWASPASI #GenAISecurity #AICTF #CTFChallenge #FinBot #GoalManipulation #AIThreats #AIAttacks #AIExploit #SecureAI #AIDefense #AIHacking #AIModelSecurity #AIAgentSecurity #AIApplicationSecurity #FinancialServicesSecurity #AdversarialAI #AIIncidentResponse #AIModelRisk #AIAppSec

InversePrompt: Turning Claude Against Itself, One Prompt at a Time - cymulate.com/blog/cve-2025-5… - By @EladBeber @ @Cymulateltd As Anthropic’s Claude Code gains traction as a powerful AI coding assistant, it promises developers a safe and streamlined way to build with Claude’s capabilities. But what happens when the same assistant meant to enforce restrictions unknowingly reveals how to bypass them? During Anthropic’s Research Preview phase, I discovered two high-severity vulnerabilities in Claude Code, which were quickly addressed by the team. These issues allowed me to escape its intended restrictions and execute unauthorized actions, all with Claude’s own help. By turning the tool inward and exploring how it interprets and validates inputs, I uncovered flaws that led to: - Path restriction bypass. - Code execution via command injection. Both are exploitable through simple prompt crafting. These findings highlight the risks of blindly trusting LLM-powered developer tools, especially when the same system meant to enforce the rules can also be used to break them. #ClaudeCode #InversePrompting #PromptInjection #LLMSecurity #AIHacking #CVE2025 #CommandInjection #PathTraversal #AIExploit #AIReverseEngineering #Anthropic #Cymulate #SecurityResearch #SandboxBypass #PrivilegeEscalation #LLMAbuse #DeveloperSecurity #SecureAI #AIHardening #ExploitResearch

GenAI Incident Response Guide - linkedin.com/feed/update/urn… The OWASP GenAI Security Project commissioned this GenAI Incident Response guide to help fill this need by providing security practitioners with guidelines and best practices for how to respond to security incidents involving GenAI applications. Authors: Bryan Nakayama, Rachel James, Keyur Rajyaguru, Madjid Nakhjiri, Rico Komenda, Waswa Mubanda, Lily R., Abhinavdutt Singh, Sarah Thornton, Volkan Kutal, Russell Tait, Hils Chan, Roddy Govender, John F., Rebekah Franolich, Sandy Dunn, Ashwed Patio, Clinton Scott, Thomas Roccia Ron F. Del Rosario, Steve Wilson, Robert Sullivan Source: genai.owasp.org/resource/gen… #GenAI #AIResponse #AISecurity #PromptInjection #ModelDrift #AIAttacks #OWASPGenAI #AIIncident #LLMVulnerabilities #AIThreatIntel #DataPoisoning #AIEthics #AISupplyChain #AIResilience #SecureAI #AIHardening #AICompliance #AIForensics #AIIncidentResponse #AIExploit @owasp #LLMSecurity

The Hidden Risk in AI-Generated Code: A Silent Backdoor A newly discovered attack method exploits AI-driven coding assistants like GitHub Copilot and Cursor, manipulating rule files to introduce silent backdoors into generated code. How the Attack Works 1️⃣ Rules File Poisoning – Attackers inject hidden malicious instructions into AI rule files, altering how code is generated. 2️⃣ Unicode Obfuscation – Invisible characters conceal harmful payloads from human reviewers but remain readable to AI models. 3️⃣ Semantic Hijacking – Subtle manipulations mislead AI models into producing insecure code, bypassing security best practices. 4️⃣ Persistent Compromise – Once a poisoned rule file enters a repository, it infects future AI-generated code, spreading via forks and dependencies. Mitigation Strategies 🔍 Audit Rule Files – Review AI configuration files for hidden Unicode characters and anomalies. 🛡 Apply AI-Specific Validation – Treat rule files with the same scrutiny as executable code. 📊 Monitor AI Outputs – Detect unexpected modifications, external dependencies, or security risks. 📖 Read more: "New Vulnerability in GitHub Copilot and Cursor: How Hackers Can Weaponize Code Agents" by Ziv Karliner, Pillar Security (@Pillar_sec). - pillar.security/blog/new-vul… #AI #CyberSecurity #AIThreats #AIBackdoor #SupplyChainSecurity #DevSecOps #MachineLearningSecurity #GitHubCopilot #CursorAI #AIHacking #SoftwareSecurity #SecureCoding #ThreatIntelligence #UnicodeObfuscation #SemanticHijacking #CyberAttack #TechRisk #AIExploit #CodeSecurity #CyberAwareness

They look like shit today, right now