GPT-4o-mini Falls for Psychological Manipulation

Interesting experiment: To design their experiment, the University of Pennsylvania researchers tested 2024's GPT-4o-mini model on two requests that it should ideally refuse: calling the user a jerk and giving directions for how to synthesize lidocaine. The researchers created experimental prompts...

Behind the Mask: Benchmarking Camouflaged Jailbreaks in Large Language Models

Large Language Models LLMs are increasingly vulnerable to a sophisticated form of adversarial prompting known as camouflaged jailbreaking. This method embeds malicious intent within seemingly benign language to evade existing safety mechanisms. Unlike overt attacks, these subtle prompts exploit...

Linux Distros Unpatched Vulnerability : CVE-2024-38274

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - Insufficient escaping of calendar event titles resulted in a stored XSS risk in the event deletion prompt. CVE-2024-38274 Note that Nessus relies on the presenc...

CVE-2025-58357

5ire is a cross-platform desktop artificial intelligence assistant and model context protocol client. Version 0.13.2 contains a vulnerability in the chat page's script gadgets that enables content injection attacks through multiple vectors: malicious prompt injection pages, compromised MCP server...

We Are Still Unable to Secure LLMs from Malicious Inputs

Nice indirect prompt injection attack: Bargury's attack starts with a poisoned document, which is shared to a potential victim's Google Drive. Bargury says a victim could have also uploaded a compromised file to their own account. It looks like an official document on company meeting policies. Bu...

CVE-2025-9190

The configuration of Cursor on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Cursor TCC Transparency, Consent, and Control permissions. Acquired resource access is limited to previously granted permission...

CVE-2025-53813

The configuration of Nozbe on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Nozbe TCC Transparency, Consent, and Control permissions. Acquired resource access is limited to previously granted permissions...

CVE-2025-9190 TCC Bypass via misconfigured Node fuses in Cursor

The configuration of Cursor on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Cursor TCC Transparency, Consent, and Control permissions. Acquired resource access is limited to previously granted permission...

CVE-2025-53811 TCC Bypass via misconfigured Node fuses in Mosh-Pro

The configuration of Mosh-Pro on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Mosh-Pro TCC Transparency, Consent, and Control permissions. Acquired resource access is limited to previously granted...

PT-2025-34759 · Cursor · Cursor

Name of the Vulnerable Software and Affected Versions: Cursor version 15.4.1 Description: The configuration of Cursor on macOS, specifically the "RunAsNode" fuse enabled, allows a local attacker with unprivileged access to execute arbitrary code that inherits Cursor TCC Transparency, Consent, and...

AI browsers could leave users penniless: A prompt injection warning

Artificial Intelligence AI browsers are gaining traction, which means we may need to start worrying about the potential dangers of something called "prompt injection." Large language models LLMs—like the ones that power AI chatbots including ChatGPT, Claude, and Gemini—are designed to follow...

Safeguarding VS Code against prompt injections

The Copilot Chat extension for VS Code has been evolving rapidly over the past few months, adding a wide range of new features. Its new agent mode lets you use multiple large language models LLMs, built-in tools, and MCP servers to write code, make commit requests, and integrate with external...

⚡ Weekly Recap: Password Manager Flaws, Apple 0-Day, Hidden AI Prompts, In-the-Wild Exploits & More

Cybersecurity today moves at the pace of global politics. A single breach can ripple across supply chains, turn a software flaw into leverage, or shift who holds the upper hand. For leaders, this means defense isn't just a matter of firewalls and patches—it's about strategy. The strongest...

Prompt-In-Content Attacks: Exploiting Uploaded Inputs to Hijack LLM Behavior

Large Language Models LLMs are widely deployed in applications that accept user-submitted content, such as uploaded documents or pasted text, for tasks like summarization and question answering. In this paper, we identify a new class of attacks, prompt in content injection, where adversarial...

Linux Distros Unpatched Vulnerability : CVE-2016-7592

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - An issue was discovered in certain Apple products. iOS before 10.2 is affected. Safari before 10.0.2 is affected. iCloud before 6.1 is affected. iTunes before...

CVE-2025-57771

Roo Code is an AI-powered autonomous coding agent that lives in users' editors. In versions prior to 3.25.5, Roo-Code fails to properly handle process substitution and single ampersand characters in the command parsing logic for auto-execute commands. If a user has enabled auto-approved execution...

CVE-2025-57771 Roo-Code potential remote code execution via auto-execute command parsing flaw

Roo Code is an AI-powered autonomous coding agent that lives in users' editors. In versions prior to 3.25.5, Roo-Code fails to properly handle process substitution and single ampersand characters in the command parsing logic for auto-execute commands. If a user has enabled auto-approved execution...

CVE-2025-57771 Roo-Code potential remote code execution via auto-execute command parsing flaw

Roo Code is an AI-powered autonomous coding agent that lives in users' editors. In versions prior to 3.25.5, Roo-Code fails to properly handle process substitution and single ampersand characters in the command parsing logic for auto-execute commands. If a user has enabled auto-approved execution...

Grok chats show up in Google searches

I’m starting to feel like a broken record, but I feel you should know that yet another AI has been found sharing private conversations so that Google was able to index them, and now they can be found in search results. It’s déjà vu in the world of AI: another day, another exposé about chatbot...

PT-2025-34449 · Robocode · Robocode

Name of the Vulnerable Software and Affected Versions: Roo Code versions prior to 3.25.5 Description: Roo Code, an AI-powered autonomous coding agent, does not correctly process process substitution and single ampersand characters within its command parsing logic for auto-execute commands. If a...