IT Security News: Afternoon roundup, 2026-11-17
Summary
Security news in this window was dominated by supply-chain and AI-adjacent risk: a fast-moving N-able zero-day, a stolen-credentials breach at an AI vendor, and a wave of unsanctioned or poorly governed AI tools across enterprises and healthcare. Several stories show attackers moving quickly from disclosure to exploitation, often within hours of a public proof of concept. Industrial control vendors also dominated the feed, with Siemens publishing a cluster of advisories tied to the Brotli "Node2Shell" flaw and several authentication and code-injection issues. For defenders, the common thread is that patching alone is not enough: shadow AI agents, exposed services, and unpatched or delayed fixes all widen the window in which attackers can act.
- Developers are deploying autonomous AI agents on local machines and unmanaged cloud instances with access to code, databases, and corporate APIs, creating "Shadow AI Agents" that operate without central oversight.
- Recorded Future surveyed 142 customers on the return on investment of threat intelligence, offering vendor-backed data points for security teams justifying budgets.
- Healthcare organizations are adopting AI tools faster than they can monitor them, and shadow AI is starting to contribute to real breaches, according to Cybersecurity Dive.
- A ClickFix social-engineering campaign using the SmartApeSG lure is delivering the CNCmachineRMS remote access trojan, as tracked by Malware-Traffic-Analysis.net.
- Public AWS Amazon Machine Images can carry malware, exposed credentials, or misconfigurations, and Qualys warns they should never be trusted based on name or description alone, referencing the whoAMI attack.
- ProjectDiscovery's httpx 1.12.0 is presented as an HTTP toolkit for probing live web servers, with guidance on distinguishing it from Kali's httpx-toolkit and Python's httpx package.
- Apple has tightened Mac permissions, and the article examines which access AI agents can still obtain despite those changes. Only the title was available, so details are limited.
- Siemens advised customers to update several industrial automation software products affected by CVE-2026-111497, a critical out-of-bounds write in the Brotli decoder bundled with Node.js, rated CVSS 9.8.
- Hackers breached an AI vendor using stolen credentials and took driver's license data belonging to Uber drivers, according to The Register.
- Attackers began exploiting the 4EverAble2Shell zero-day in N-able N-central within hours of a public proof of concept, gaining root-level code execution even on fully patched servers with the Legacy API enabled.
- Microsoft Security Blog research shows device code phishing can extend to AI coding agents, using GitHub Copilot CLI as an example of a technique that affects any device-flow client.
- Siemens ROS# before version 2.3.1 has a code injection flaw (CVE-2026-109097) that can let an attacker with a crafted message definition file inject arbitrary code into generated programs, and Siemens advises updating.
- Researchers at Physics-focused ScienceDaily report that laser flashes can align tiny electrical regions in a crystal for about a trillionth of a second, a possible step toward new information processing methods. This is basic research, not a security threat.
- Siemens software products including Teamcenter, Opcenter, and Xpedition are listed as affected by the Brotli CVE-2026-111497 issue, with Siemens pointing to separate field bulletins for several product lines.
- Microsoft traces a global Microsoft 365 outage to a months-old Windows defect and to recovery systems that spread the failure, raising concerns about delivery of critical security alerts, per eSecurityPlanet.
- Siemens SIMATIC products, including WinCC, PCS 7, and Process Historian, are affected by CVE-2026-58114, a weak NTLM authentication flaw that allows machine-in-the-middle relay attacks, and Siemens has no fix planned for some versions.
- Huntress found attackers exploiting GeoVision camera software at 13 organizations, targeting Windows security tools and installing the XMRig cryptominer.
- Dorsett Controls InfoScan versions up to 1.45.0 contain five vulnerabilities, including improper certificate validation and hard-coded cryptographic keys, and the vendor recommends upgrading to version 1.46 or later.
- A financially motivated group running the SANSAR toolkit hijacked exposed Ray job agents on Linux GPU servers, using Tor, Ethereum blockchain lookups, and falsified GPU readings to hide its cryptomining, according to Elastic Security Labs.
- Siemens meta-iot2050 firmware V01.03.01 and later contains a command injection flaw (CVE-2026-109830) in the firmware update tool, and no fix is currently available.
- Huntress analyzed phishing kits EvilTokens, Knight Office, and Kali365, which steal and replay Microsoft 365 session tokens, and outlined detection and containment steps for defenders.
- Siemens SIMATIC WinCC V7.5, V8.0, and V8.1 Graphics Designer can execute active content when files are opened (CVE-2026-59082), and no fix is available yet, so Siemens advises opening only trusted files.
- Black Duck's 2027 AI Readiness Report analyzed 65.5 million application security findings and argues that severity scores alone fail teams, so backlogs keep growing even as fixing effort rises.
- Siemens Mendix Studio Pro versions 11.12 and 11 contain CSRF and unrestricted-binding flaws in the MCP Server (CVE-2026-112136, CVE-2026-112137) that could allow remote data access or tool execution, fixed in 11.12.7 and 11.16.0.
Summaries written with AI from the linked source articles.
Sources in this roundup
| All CISA Advisories |
|
8 article(s) |
| esecurityplanet |
|
3 article(s) |
| CySecurity News – Latest Information Security and Hacking Incidents |
|
2 article(s) |
| Microsoft Security Blog |
|
2 article(s) |
| BleepingComputer |
|
1 article(s) |
| Cybersecurity Dive – Latest News |
|
1 article(s) |
| DZone Security Zone |
|
1 article(s) |
| Darknet – Hacking Tools, Hacker News & Cyber Security |
|
1 article(s) |
| Hacking News — ScienceDaily |
|
1 article(s) |
| IT SECURITY GURU |
|
1 article(s) |
| Malware-Traffic-Analysis.net – Blog Entries |
|
1 article(s) |
| Qualys Security Blog |
|
1 article(s) |
| Recorded Future |
|
1 article(s) |
| www.theregister.com – Articles |
|
1 article(s) |
Most-mentioned keywords
| siemens |
|
7 mention(s) |
| agents |
|
3 mention(s) |
| hackers |
|
3 mention(s) |
| risks |
|
3 mention(s) |
| authentication |
|
2 mention(s) |
| code |
|
2 mention(s) |
| coding |
|
2 mention(s) |
| data |
|
2 mention(s) |
Sources
- The Rise of the Unsanctioned Agent Swarm
- What Is Threat Intelligence Worth? 142 Security Teams Did the Math.
- Healthcare firms fail to monitor AI tools, even as shadow AI begins leading to breaches
- 2026-11-12: SmartApeSG ClickFix campaign pushes CNCmachineRMS RAT
- Understanding the Risks of Public AWS AMIs: Hidden Dangers of Malicious and Misconfigured Images
- httpx – ProjectDiscovery’s HTTP Toolkit for Probing Live Hosts
- Apple Tightens Mac Permissions: What AI Agents Can Still Access
- Siemens Industrial Automation Software Products
- Hackers smash 'n' grab AI vendor: Take off with license data of Uber drivers
- Hackers Exploit New N-able N-central Zero-Day Hours After PoC Release
- The risks of device code authentication for AI coding agents
- The risks of device code authentication for AI coding agents
- Siemens ROS#
- Physicists use laser light to unlock a crystal's hidden electrical powers
- Siemens Industrial Software Products
- Microsoft 365 Outage: How Recovery Systems Spread the Failure
- Siemens SIMATIC
- Hackers Target GeoVision Systems at 13 Organizations, Huntress Finds
- Dorsett Controls InfoScan
- New Linux Malware Uses Tor, Ethereum Blockchain, and Falsified GPU Data to Run Cryptomining Operations Undetected
- Siemens meta-iot2050
- Inside Modern Phishing Kits Built to Steal Microsoft 365 Sessions
- Siemens SIMATIC WinCC
- Fix what matters, not what’s easy: why severity scores are failing security teams
- Siemens Mendix Studio Pro
