Advertisement
AI-Powered Vulnerability Discovery: Automated Zero-Day Mining with LLMs
Intruder reveals an AI-driven system combining code slicing and LLMs to automatically discover complex software vulnerabilities and zero-day exploits.
Frontier AI Governance: Managing Cybersecurity Risks of Autonomous Models
Explore the cybersecurity challenges and governance needs of increasingly autonomous AI models, as states seek transparency amidst emerging risks.
xAI Grok Build Repository Upload Risks: Analyzing CLI Data Exposure
xAI's Grok Build CLI version 0.2.93 discovered uploading entire Git repositories, including history and secrets, to remote storage without user consent.
Yellow Teams: Pioneering Adversarial AI Security Methodologies
Explore how 'Yellow Teams' are crucial for assessing AI security, developing defensive strategies, and understanding AI's potential as both a cyber weapon and a shield.
MemGhost Attack: Persistent Memory Poisoning in AI Agents via Email
The MemGhost attack targets AI agent memory systems via email, planting persistent false facts to stealthily manipulate long-term assistant behavior.
Scans Target Model Context Protocol Servers and AI Credentials
Security researchers observe an increase in scans targeting Model Context Protocol (MCP) servers and AI credentials, potentially exposing sensitive data.
Advertisement
Anthropic Extends Claude Fable 5 Access: Enterprise Migration Impacts
Anthropic extends Claude Fable 5 access for paid users until July 19. Learn how this availability extension impacts enterprise AI security and infrastructure.
Ghostcommit: Hidden Prompt Injection in Images Targets AI Agents
Researchers demonstrate Ghostcommit, a technique using images to hide prompt injection attacks that trick AI agents into exfiltrating repository secrets.
UK's Agentic AI Cyber Defense Plan & Industry Pledge
The UK government unveils its Agentic AI Cyber Defense Plan and an industry pledge involving 16 tech giants to bolster national security and cyber resilience.
AI Coding Agents Vulnerable to Friendly Fire Command Execution
AI coding agents like Anthropic's Claude Code and OpenAI's Codex are susceptible to Friendly Fire attacks, leading to unintended local code execution.
Defensive AI Agents: Countering the Rise of Local AI Model Attacks
The true AI threat is not large frontier models, but cheap local AI models enabling scalable attacks. Learn why CISOs must build defensive AI agents now.
CVE-2024-37014: CISA Orders Federal Agencies to Patch Langflow
CISA added CVE-2024-37014, a critical authentication bypass in the Langflow AI framework, to its KEV catalog following reports of active exploitation.
Writer AI Platform: Critical Session Isolation Flaw 'WriteOut'
Runtime Rebel details the critical 'WriteOut' session isolation vulnerability in Writer AI, enabling cross-tenant compromise and unauthorized agent takeover.
Keyfactor's $1B+ Investment: Addressing AI & Post-Quantum Threats
Keyfactor secures over $1 billion to advance its machine identity and PKI platform, preparing enterprises for future AI-driven and post-quantum cryptographic challenges.
JadePuffer: First LLM-Driven Ransomware Leverages Langflow Flaw
Analysis of JadePuffer, the first reported LLM-driven ransomware, which exploited a Langflow vulnerability to exfiltrate database data and encrypt systems.
SkillCloak: Malicious AI Agent Skills Evade Static Code Scanners
Researchers at HKUST reveal SkillCloak, a technique using self-extracting packing to allow malicious AI agent skills to bypass static security analysis.
Agentic AI Automates Ransomware Attacks via Langflow Exploitation
Agentic AI agents demonstrate automated multi-stage ransomware attacks using Langflow, raising concerns for AI development security and future threat automation.
Auditing AI-Driven Software Development: Security Governance Strategies
Learn how to audit AI-generated code and govern AI tool usage to mitigate security risks in modern software development lifecycles.
AI Compute Hijacking and BlueHammer Ransomware Analysis
Analysis of emerging threats including AI compute hijacking via sandbox escapes, BlueHammer ransomware TTPs, and logic flaws in Apple email services.
How Agentjacking Exploits AI Coding Agents via Fake Bug Reports
Researchers demonstrate 'Agentjacking,' a technique using indirect prompt injection to hijack AI coding agents through malicious GitHub bug reports.
Anthropic Claude Fable 5 Restored After U.S. Export Control Lift
Anthropic restores global access to Claude Fable 5 following the lifting of U.S. export controls linked to jailbreaking vulnerabilities and security concerns.
Phantom Squatting: How Attackers Weaponize AI-Hallucinated Domains
Security researchers have identified 'Phantom Squatting,' a technique where attackers register non-existent domains hallucinated by AI for phishing.
AI-Generated Workflows: Hidden Vulnerabilities & Control Gaps
Explore the silent security disaster of AI-generated workflows. Understand hidden vulnerabilities, control gaps, and the critical need for human oversight in AI-driven…
Claude Code Indirect Prompt Injection: Hijacking Developer Machines
Researchers demonstrate a new attack method leveraging indirect prompt injection in Claude Code, enabling the hijack of developer machines via malicious code in…