In this week's Seclog, a critical theme emerging is the escalating security challenges posed by Artificial Intelligence, with multiple reports detailing vulnerabilities in AI assistants, social networks, and even children's toys, alongside the intriguing development of AI autonomously discovering zero-day exploits. The landscape is further complicated by significant supply chain and critical infrastructure compromises, including state-sponsored hijacking of a popular editor and severe RCE vulnerabilities in enterprise platforms like Samsung MagicINFO, Google Cloud's Apigee, and Kubernetes. Attackers continue to leverage sophisticated tactics, from one-click RCEs to exploiting authentication bypasses in widely used systems like Teleport, emphasizing the persistent need for robust security postures. Meanwhile, new botnets like Badbox 2.0 highlight the ongoing threat from malicious infrastructure, while the community actively develops tools for offensive capabilities, such as browser data exfiltration, and defensive measures, like Python wheel scanners. The reports collectively underscore a rapidly evolving threat environment where AI plays a dual role in both creating new attack surfaces and potentially aiding in their discovery.

📚 SecMisc #

ProjectDiscovery Launches OSS Bounty Program - projectdiscovery.io

ProjectDiscovery launched an open-source software bounty program to incentivize security research and contributions to their tools. This initiative aims to improve the security posture of widely used open-source security tools by rewarding external vulnerability discoveries.

PortSwigger's Top Web Hacking 2025 - portswigger.net

PortSwigger's annual community-driven report identifies the most innovative and impactful web hacking techniques from 2025. This resource provides a vital overview of emerging attack vectors and advanced methodologies for web security professionals.

Sandboxed, Persistent Browser Setup - tech.michaelaltfield.net

A guide on configuring a persistent, single-site browser isolated with firejail and routed through proxychains for a static IP. This setup enhances operational security by providing a dedicated, sandboxed environment for specific tasks, preventing host system contamination and ensuring network anonymity.

Balancing AI, Security, and Productivity - pentesterlab.com

This blog post discusses the perpetual challenge of balancing security and privacy with productivity, especially in the context of integrating AI. It explores how maintaining high security standards, like air-gapped environments, often comes at a significant cost in time and resources, a trade-off magnified by AI adoption.

📰 SecLinks #

State-Sponsored Hackers Hijack Notepad++ - notepad-plus-plus.org

A popular open-source text editor, Notepad++, reported a security incident involving a hijack by state-sponsored actors. This incident, stemming from a shared hosting provider compromise, underscores the critical supply chain risks affecting even widely used software.

1-Click RCE Hacks AI Assistant - depthfirst.com

A critical 1-Click Remote Code Execution (RCE) vulnerability was discovered in OpenClaw (Moltbot/ClawdBot), an AI personal assistant used by over 100,000 developers. This exploit, chaining multiple findings, allowed attackers to gain unrestricted local computer control and access sensitive data/keys by merely visiting a malicious webpage.

Instagram Vulnerability Exposes Private Posts - 003random.com

A researcher discovered a vulnerability in Instagram that allowed bypassing privacy settings to view posts from private accounts. This write-up details the method used, highlighting how user interaction (blocking) could be leveraged to subvert core privacy controls.

AI Discovers Node.js, React Zero-Days - winfunc.com

An AI system autonomously discovered zero-day vulnerabilities in Node.js and React, two foundational JavaScript runtimes and frameworks. This marks a significant advancement in automated vulnerability research, indicating AI's growing capability to identify critical flaws in widely deployed software.

Exploiting Teleport Authentication Bypass CVE - blog.offensive.af

Detailed analysis and exploitation of CVE-2025-49825, an authentication bypass vulnerability in Teleport. The post provides insights into the root cause and practical exploitation techniques for this critical security flaw in target infrastructures.

Moltbook AI Network Exposed API Keys - wiz.io

Researchers uncovered a critical vulnerability in Moltbook, an AI social network, leading to the exposure of millions of API keys. The incident highlights severe security control deficiencies in rapidly developed AI applications, particularly those handling sensitive access tokens.

Badbox 2.0 Botnet Operators Identified - krebsonsecurity.com

Investigation into the Badbox 2.0 botnet, which infects millions of Android TV streaming boxes, reveals potential operators. The compromise of Badbox 2.0's control panel by the Kimwolf botnet suggests inter-criminal rivalries and provides insights into attribution efforts.

Samsung MagicINFO 9 RCE Chain - srcincite.io

Researchers uncovered a chain of multiple vulnerabilities in Samsung MagicINFO 9, leading to pre-authenticated Remote Code Execution. This detailed write-up demonstrates bypassing patch attempts and chaining flaws to achieve critical impact on digital signage infrastructure.

Hacking AI Children's Toy for Access - josephthacker.com

A vulnerability was discovered in a new AI children's toy, enabling remote access to every conversation the child has with the device. This highlights significant privacy and security concerns in consumer-grade AI products, especially those interacting with children.

GCP Apigee Cross-Tenant Vulnerability - omeramiad.com

A critical cross-tenant vulnerability (CVE-2025-13292) named "GatewayToHeaven" was found in Google Cloud's Apigee. This flaw allowed reading and writing verbose cross-tenant access logs containing plaintext access tokens, enabling potential impersonation of any end user across organizations.

🐦 SecX #

WhatsApp Fingerprinting Reveals Moltbot Users - x.com

A security researcher developed a WhatsApp fingerprinting tool capable of identifying users linked to OpenClaw/Moltbot/ClawdBot AI assistants. This technique demonstrates a novel way to track user connections to third-party AI agents via communication platforms.

Kubernetes RCE Via Read-Only RBAC - x.com

Research disclosed a critical Remote Code Execution (RCE) vulnerability in Kubernetes, allowing arbitrary command execution across all pods with "read-only" RBAC permissions. The vulnerability bypasses logging mechanisms and enables trivial pod breakout, posing a significant unpatched threat to cluster security.

🎥 SecVideo #

Jim Manico on AI Security - youtube.com

A video recording of Jim Manico's talk at NDC AI 2025, providing an introduction to AI security. This presentation likely covers fundamental concepts, common threats, and defensive strategies pertinent to securing AI systems and applications.

💻 SecGit #

Claude AI Code RCE Proof - github.com

This GitHub repository provides a Proof-of-Concept (PoC) for Remote Code Execution in Claude AI code. The PoC demonstrates critical vulnerabilities in AI agent environments, showcasing the potential for arbitrary command execution.

Dump Browser Secrets Tool - github.com

A repository for DumpBrowserSecrets, a tool designed to extract sensitive data from modern Chromium and Gecko-based browsers. This utility can exfiltrate refresh tokens, cookies, saved credentials, credit cards, and browsing history, proving valuable for post-exploitation scenarios.

AI-Assisted Android Frida Instrumentation - github.com

A GitHub project for kahlo-mcp, a Frida Mobile Code Platform (MCP) server designed to integrate autonomous AI assistance for Android instrumentation. This tool aims to enhance dynamic analysis and reverse engineering of Android applications by leveraging AI for more efficient hook development and analysis.

Python Wheel File Security Scanner - github.com

wheelaudit is an open-source Python tool for scanning .whl files to detect security issues prior to installation. It identifies threats like path traversal (CVE-2026-24049), RECORD tampering, command shadowing, embedded secrets, and zip bombs, enhancing supply chain security for Python projects.

← Back to Seclog