<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Keryn Headlines: Daily edition</title><link>https://keryn.dev/headlines/daily/</link><description>The daily edition of Keryn Headlines: application and AI security for teams that build with AI.</description><language>en</language><lastBuildDate>Sun, 22 Nov 2026 15:38:39 GMT</lastBuildDate><atom:link href="https://keryn.dev/headlines/daily/feed.xml" rel="self" type="application/rss+xml"/><item><title>Daily, Sunday 22 November 2026 (No. 6)</title><link>https://keryn.dev/headlines/daily/2026-11-22/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-22/</guid><pubDate>Sun, 22 Nov 2026 06:00:00 GMT</pubDate><description>If you run the MONAI Label imaging server: Amazon Threat Intelligence said suspected China-linked hackers used autonomous AI agents to find a previously unknown flaw in it and gain root access to a university server . No source states a fixed build yet . Our advice: keep it off untrusted networks until an advisory appears. If your applications use GPGME: the GnuPG project warned that a malicious OpenPGP key with an unusually long user ID can trigger a critical memory bug that attackers could use to steal application data . The project has released something, but the excerpt ends before naming it . Our advice: check GnuPG&#39;s announcement and treat inbound keys as untrusted meanwhile. If you run clusters on Rackspace Spot: Rackspace warned that hackers used compromised credentials to deploy unauthorized workloads and access Kubernetes secrets in customer clusters . The report does not say how the attackers got the credentials . Our advice: rotate those secrets and review unfamiliar workloads. Top: Attackers&#39; AI Agents Found a Zero-Day in Medical-Imaging Software and Gained Root Access | GnuPG Warns Critical GPGME Flaw Could Expose App Data via Malicious OpenPGP Keys | Rackspace Spot breach let hackers access customer Kubernetes secrets</description></item><item><title>Daily, Saturday 21 November 2026 (No. 5)</title><link>https://keryn.dev/headlines/daily/2026-11-21/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-21/</guid><pubDate>Sat, 21 Nov 2026 06:00:00 GMT</pubDate><description>If you run F5 BIG-IP, F5 has released emergency patches and warns customers to patch five critical vulnerabilities that attackers can exploit for denial-of-service or remote code execution . Our sources name no fixed build, so take build numbers from F5&#39;s own advisory before you patch. If you run PHP with its default session handler, Cloudflare says the phpAfterShell flaw can lead to code execution . Cloudflare says its WAF blocks it for all customers, including Free plans, and adds rules for other PHP flaws that the same PHP release fixes . Our sources do not name that release, so find it and upgrade rather than rely on the WAF alone. If you still run SonicWall Generation 6 firewalls, SonicWall says they will not get security patches, even for critical flaws . It urges customers to retire them immediately and stop exposing them on the internet edge, citing risks from AI-assisted attacks . Until you replace them, we suggest you treat any internet-facing Gen 6 device as unpatchable. Top: Amazon says Chinese hackers used Cairn AI tool to discover apparent zero-day during attack on university | F5 releases emergency patches for critical BIG-IP RCE flaws | How Cloudflare WAF helps protect against phpAfterShell and other critical PHP flaws</description></item><item><title>Daily, Friday 20 November 2026 (No. 4)</title><link>https://keryn.dev/headlines/daily/2026-11-20/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-20/</guid><pubDate>Fri, 20 Nov 2026 06:00:00 GMT</pubDate><description>If you run BerriAI LiteLLM: CISA on Wednesday added a LiteLLM vulnerability to its Known Exploited list, alongside flaws in Synology DSM, Apache Axis and REDCap . The excerpt does not name the LiteLLM CVE or say how it was exploited, and no fixed build is stated yet . Find your LiteLLM instances and check the vendor advisory. If you run VMware vCenter Server: Broadcom has released security updates for a critical authentication bypass . BleepingComputer reports that Chinese-linked hackers have exploited it in zero-day attacks since September 2026 . The excerpt does not name the fixed build or give indicators of compromise . Apply Broadcom&#39;s update, and watch for its guidance on what to check. If staff build AI agents in project trackers, CRMs or automation tools: Nudge Security reports that these agents often run on their creator&#39;s credentials and carry that person&#39;s access to company data . The excerpt does not describe the sample . Inventory these agents and record whose access each one uses. Top: CISA Warns of 4 Exploited Flaws Linked to Ransomware, Cryptomining, and LLMjacking Attacks | VMware patches vCenter zero-day exploited in attacks since September | PHP Fixes 7 Critical Flaws, Including Bugs That Could Allow Remote Code Execution | Ruby2Shell: critical Ruby RCE, affected versions and how to fix it (GHSA-34m4-m9fh-gqp5) | Drupal patches critical RCE flaw, warns exploits may surface quickly</description></item><item><title>Daily, Thursday 19 November 2026 (No. 3)</title><link>https://keryn.dev/headlines/daily/2026-11-19/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-19/</guid><pubDate>Thu, 19 Nov 2026 06:00:00 GMT</pubDate><description>If you run NGINX with HTTP/2, Wiz reports two critical flaws, CVE-2026-123633 and CVE-2026-123007, that could allow unauthenticated remote code execution . Vendor records list Red Hat Enterprise Linux 8, 9 and 10 and Red Hat Hardened Images as affected by CVE-2026-123007, with no fixed build stated yet . Find your HTTP/2 listeners and watch for vendor fixes. The source does not say whether either flaw is exploited . If you run Apache Roller, Apache warns that threat actors are exploiting a critical flaw to gain code execution, and it has released security updates . The source names neither the affected versions nor the fixed build, so check Apache&#39;s advisory before calling a server patched . If you use Ubuntu Pro, Canonical lets you install security patches written by AI agents before its engineers finish reviewing and testing them . The patches come from Surge, a Canonical system for fixing known vulnerabilities in Ubuntu&#39;s userspace software . Keep these patches out of production until review is complete. Top: Unauthenticated Remote Code Execution in NGINX: CVE-2026-123633 and CVE-2026-123007 | Apache warns of Roller RCE vulnerability exploited in attacks | CVE-2026-90817: Vanderbilt University Research Electronic Data Capture (REDCap) Code Injection Vulnerability added to CISA&#39;s Known Exploited Vulnerabilities catalog</description></item><item><title>Daily, Wednesday 18 November 2026 (No. 2)</title><link>https://keryn.dev/headlines/daily/2026-11-18/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-18/</guid><pubDate>Wed, 18 Nov 2026 06:00:00 GMT</pubDate><description>If you install Microsoft&#39;s PyPI package agent-framework, check locked and installed versions for 1.26.1 and 1.26.2 . Wiz reports both versions shipped with an added credential stealer on November 17, 2026 at 20:35 UTC, and now sit in quarantine . The package gets almost 200k downloads a month, and Wiz&#39;s headline names Azure and CI pipelines as targets . Wiz&#39;s excerpt does not list which credentials the stealer takes . Our advice is to rotate any secret an environment running either version could reach. If you build or run Java services on OpenJDK, track CVE-2026-114938, a remote code execution flaw Datadog calls Java4Shell . Datadog&#39;s excerpt names no affected versions or fix . We suggest you inventory OpenJDK use and watch for fixes. If your coding agents sign in through device code flow, note that Microsoft presents this as lab research, not a reported attack . Microsoft shows how device code phishing can extend to AI coding agents, using GitHub Copilot CLI as an example . We suggest you review which agents accept device codes. Top: Java4Shell: Remote Code Execution in OpenJDK (CVE-2026-114938) | Microsoft agent-framework Compromised: Credential Stealer Targets Azure and CI Pipelines | Optimizely supply chain attack served malware on CNN, Fox News, BBC sites | N-able Patches 17 N-central Flaws as Attackers Exploit Earlier Bugs to Deploy Crypto Miners | Cloud keys leaked in Docker Hub images are stolen within seconds</description></item><item><title>Daily, Tuesday 17 November 2026 (No. 1)</title><link>https://keryn.dev/headlines/daily/2026-11-17/</link><guid isPermaLink="true">https://keryn.dev/headlines/daily/2026-11-17/</guid><pubDate>Tue, 17 Nov 2026 06:00:00 GMT</pubDate><description>If your projects depend on WPPConnect from npm, look for releases 2.4.1 and 2.10.82 . StepSecurity reports they are malicious, shipped through the project&#39;s own pipeline, and steal secrets and WhatsApp sessions via Telegram . Remove them and rotate what they could reach . If your developers use Claude Code extensions, SecurityWeek reports that an attacker stole a Fireworks AI access key . The attacker then exfiltrated credentials that let them push malicious copies of a Claude Code extension . Check each extension&#39;s publisher and rotate tokens wherever a copy ran. The excerpt names no affected versions . If you expose LiteLLM, One API or Dify to the internet, Amazon&#39;s threat intelligence team says attackers&#39; AI agents are hunting these apps for stored AI provider keys . The evidence comes from the agents&#39; own records, which Amazon captured with decoys . The excerpt does not say how many real deployments lost keys . Put these apps behind authentication and watch provider key usage. Top: Cursor Supply Chain Attack Started at Fireworks AI | WPPConnect Compromised: Malicious Releases Steal Secrets and WhatsApp Sessions Through the Project&#39;s Own Pipeline | California sues OpenAI over AI agents escaping sandboxes, hacking firms | Gemini AI agents repeatedly bypassed safeguards to attack real website | China-Nexus Actors Breach 10 Organizations in AI-Driven Attacks</description></item></channel></rss>