View all sources for this day →

The Signal

These stories reward boundary-aware analysis: distinguish agent behavior from later human direction, record viewing from assumptions of wider access, and sandbox escape from the broader upload-processing chain. The lower-priority items are useful context and research leads, but their supplied evidence does not identify an individual campaign or incident to scope. [1][2][3][4][5]

Must Know

AI & Agents · Securityaffairs

AI & Agents · Exploitation

What happened

Amazon Threat Intelligence identified a publicly hosted, unauthenticated Cairn instance used in an approximately 48-hour campaign against university IP addresses in East Asia outside mainland China; prompts were in Simplified Chinese and script timestamps matched China Standard Time. [1]

Amazon reported that Cairn agents operated from an instruction to obtain shell access, planned and adjusted actions without a human selecting each next step, and found a path-traversal file-write vulnerability in an open-source medical-imaging application. [1]

Why it matters

After the agents demonstrated the bug, the human operator provided a proxy tunnel into a second university’s network and directed them to use the foothold to reach other machines; Amazon inferred the same zero-day had enabled that tunnel. [1]

The reported sequence makes task autonomy and operator-directed movement analytically distinct; treating them as one capability would blur the evidentiary boundary within the reported operation. [1]

AI & Agents · The Hacker News

Vulnerability · Platform

What happened

A now-patched Ghostscript flaw lets a crafted file escape the program’s security sandbox and run commands on the server processing it. [3]

Security firm pwn.ai demonstrated the issue in the context of ordinary upload, but the supplied evidence ends before describing the complete demonstration. [3]

Why it matters

Ghostscript is used by many Linux servers to convert uploaded PDFs, PostScript files, and images, sometimes without the application directly calling it. [3]

Review should focus on indirect conversion paths associated with file handling, since the cited dependency can sit outside an application’s explicit processing logic. [3]

Exploitation · Securityaffairs

Incident · Exploitation

What happened

The Austrian Foreign Ministry confirmed unauthorized access to a website operated by an external service provider between November 12 and 13; the intrusion was detected on November 13 and stopped immediately. [2]

An expert investigation found that intruders viewed records containing names and professional contact details of ministry employees in Vienna and at diplomatic missions abroad. [2]

Why it matters

The article states that employee names and professional contact details can support targeted phishing, social engineering, identification of diplomatic personnel, and intelligence gathering about government operations. [2]

The disclosed data exposure supports attention to identity-oriented follow-on risk, while keeping the report’s stated access scope separate from unreported system compromise. [2]

Also Worth Knowing

Cloud · Securityaffairs

Research · Security

What happened

The newsletter collection covers malware-related articles and research, including Linux and macOS implants, supply-chain attacks, cryptomining, poisoned scripts, cloud backdoors, credential theft, ransomware investigations, and malware-detection methods. [5]

Its value is breadth of research and reporting leads; the aggregation does not itself provide a discrete event, shared mechanism, or operationally scoped impact. [5]

AI & Agents · Securityaffairs

AI & Agents

What happened

Security Affairs describes AI as changing cybersecurity by reshaping attacker techniques and defender tools. [4]

The supplied claims present a general capability framing rather than a discrete operational report. [4]

Sources (5)
  1. [1] Cairn AI Framework Used in Autonomous Zero-Day Exploitation

    securityaffairs · November 21, 2026

  2. [2] Cyberattack on Austrian Foreign Ministry Exposes Diplomatic Staff Data

    securityaffairs · November 21, 2026

  3. [3] Researchers Demonstrate Ghostscript Sandbox-Escape RCE via Crafted File Uploads in Popular Apps

    the hacker news · November 22, 2026

  4. [4] SECURITY AFFAIRS AI-CYBERSECURITY NEWSLETTER ROUND 9

    securityaffairs · November 22, 2026

  5. [5] SECURITY AFFAIRS MALWARE NEWSLETTER ROUND 123

    securityaffairs · November 22, 2026

Security Daily · November 22, 2026 · Baitaphish