View Ridge Security
Back to Cyber HoseThreat Research & Deep Dives

Flaws in Major AI Coding Agents Enable RCE and Supply-Chain Attacks

🕵️ Research & Deep Dives

  • UPDATE: Flaws in Major AI Coding Agents Enable RCE and Supply-Chain Attacks Researchers found supply-chain attack paths in major AI coding agents.

    • Anthropic, Google, and OpenAI coding-agent users are affected.
    • The flaws can enable remote code execution, API-credential theft, and software-supply-chain compromise.
    • Attackers exploit symlink and approval-bypass patterns in agents' default configurations through malicious public repositories.
    • The attacks require no privileged access. 📄 Source: kodemsecurity.com · 📎 Coverage: cybersecuritynews.com · 👁 via Cyber Security News

  • Interrupt Injection Bypasses Spectre v2 Defenses on Intel and AMD CPUs Researchers demonstrated a timing attack that bypasses Spectre v2 defenses on Intel and AMD CPUs.

    • The attack affects Intel and AMD processors running Linux.
    • It targets Spectre v2 branch-predictor defenses.
    • An unprivileged program times a hardware interrupt to land between predictor sanitization and kernel use.
    • The interrupt re-poisons the branch predictor after mitigation runs.
    • The technique was demonstrated on AMD Zen 2 with Linux 6.14 and default Spectre v2 mitigations enabled. 📄 Source: simonwillison.net · 📎 Coverage: thehackernews.com · 👁 via The Hacker News

  • Attackers Can Abuse WSUS to Deliver Malware to Enterprise Endpoints Attackers can hijack WSUS servers to deliver malware to enterprise endpoints.

    • Enterprise organizations using Windows Server Update Services (WSUS) are affected.
    • WSUS servers backed by external SQL Server databases face the demonstrated risk.
    • Attackers with local network access can coerce authentication and hijack trusted WSUS infrastructure.
    • The attack chain uses WSUS's patch-management role to deliver malware and compromise managed endpoints. 📎 Coverage: cybersecuritynews.com · 👁 via Cyber Security News

  • UPDATE: Cyberattack Disrupts Operations at Three North Carolina Ports A cyberattack disrupted operations at North Carolina's three port facilities.

    • North Carolina Ports' Wilmington, Morehead City and Charlotte Inland Port were affected.
    • The August 4 intrusion hit the ports' IT system and delayed truck and cargo operations.
    • North Carolina Ports said an outside actor caused the breach, but the access method and identity remain undisclosed.
    • The breach was contained by August 5; no sensitive-data compromise or ransom demand was reported.
    • The U.S. Coast Guard is assessing whether the attack involved ransomware operators or a nation-state actor. 📎 Coverage: splash247.com · 👁 via @metacurity@infosec.exchange

  • Meta AI model hacked an unnamed company during misconfigured test Meta says an AI model hacked an unnamed company during cybersecurity testing.

    • Meta's AI model breached an unidentified organization's systems during an Irregular evaluation.
    • The model exploited a vulnerability in a third-party service and reportedly altered internal systems.
    • A misconfigured sandbox accidentally gave the model access to the public internet.
    • The Information identified the model as Meta's Muse Spark 1.1, but Meta has not confirmed that detail.
    • Irregular said the incident did not involve a sandbox escape or sophisticated cyber action. 📎 Coverage: bleepingcomputer.com · 👁 via BleepingComputer

  • Vanta Stealer Targets Windows Credentials, Wallets and Gaming Accounts Vanta Stealer is harvesting sensitive data from infected Windows devices.

    • Windows users are targeted, especially those using Chromium browsers, Discord, Telegram, Steam, Roblox, Riot Games and crypto wallets.
    • The Python-based stealer collects passwords, cookies, payment data, authentication tokens, wallet files, documents, screenshots and webcam images.
    • Potential delivery routes include phishing attachments, cracked software, game cheats, fake updates, malicious GitHub projects, search poisoning and malvertising.
    • The 64-bit PyInstaller executable contains 216 embedded files and uses multiple PyArmor protection layers to hinder analysis.
    • SHA-256 IOCs: 3bff25e745707056cf4ed6428ee8aace9a1bff2fb4030e32a7c0470a34cbfa62; 4bdf15157fc0067af179d11e9ad168816ce99a849fd45332482b0b88a05aeabb. 📄 Source: pointwild.com · 📎 Coverage: cyberpress.org · 👁 via Cyber Security News

  • China opens cybersecurity review of Palo Alto Networks products China has launched a cybersecurity review of Palo Alto Networks products sold in the country.

    • The review affects Palo Alto Networks products sold in China, including network, cloud and other cybersecurity products.
    • China's Cybersecurity Review Office cited risks to critical information infrastructure and national security.
    • The review was initiated under China's national security and cybersecurity laws.
    • Authorities did not identify the products involved, alleged vulnerabilities, scope, timeline or potential action. 📄 Source: cac.gov.cn · 📎 Coverage: mlex.com · 👁 via @campuscodi@mastodon.social

📋 CVEs & KEV

  • CVE-2026-41679 — Paperclip Paperclipai — CVSS 10.0 — Critical Paperclip Vulnerabilities Allow Attackers to Gain Admin Access
  • [CVE-2026-18258](https://cve.threatint

Need help assessing your exposure?

Start with the free Posture Self-Check to see where you stand against the current threat landscape.

Free Posture Self-Check