PREVIEW. This is how it will look in an inbox; it was not sent to you. Each real email carries its own unsubscribe link and Keryn's postal address. KERYN HEADLINES Daily, Wednesday 18 November 2026 No. 2 Covers 17 November 2026 00:00 to 18 November 2026 00:00 UTC Web: https://keryn.dev/headlines/daily/2026-11-18/ FROM THE EDITOR If you install Microsoft's PyPI package agent-framework, check locked and installed versions for 1.26.1 and 1.26.2 [2]. Wiz reports both versions shipped with an added credential stealer on November 17, 2026 at 20:35 UTC, and now sit in quarantine [2]. The package gets almost 200k downloads a month, and Wiz's headline names Azure and CI pipelines as targets [2]. Wiz's excerpt does not list which credentials the stealer takes [2]. 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 [1]. Datadog's excerpt names no affected versions or fix [1]. 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 [6]. Microsoft shows how device code phishing can extend to AI coding agents, using GitHub Copilot CLI as an example [6]. We suggest you review which agents accept device codes. AI-drafted. Not reviewed by a human editor: it passed automated checks only. Published by Keryn. Corrections: keryn@keryn.dev TOP DEVELOPMENTS [1] Java4Shell: Remote Code Execution in OpenJDK (CVE-2026-114938) Datadog · Advisory · https://securitylabs.datadoghq.com/articles/java4shell-remote-code-execution-in-openjdk-cve-2026-114938 Why it matters: A remote code execution flaw dubbed Java4Shell (CVE-2026-114938) in OpenJDK may affect any Java-based service your team builds or runs, so check whether you use OpenJDK and watch for fixes. Builds (CVE-2026-114938, from vendor records): fixed in Amazon Corretto 27 27.0.1.1.4, Red Hat Build of OpenJDK 17.0.21.1 , Red Hat Build of OpenJDK 21.0.13.1 , Red Hat Build of OpenJDK 25.0.5.1 , Temurin Build of OpenJDK 25.0.5.1 ; affected, no fix stated: Red Hat build of Keycloak. [2] Microsoft agent-framework Compromised: Credential Stealer Targets Azure and CI Pipelines Wiz · Incident · https://wiz.io/blog/malicious-version-of-microsofts-agent-framework-published-to-pypi [3] Optimizely supply chain attack served malware on CNN, Fox News, BBC sites BleepingComputer · Incident · https://bleepingcomputer.com/news/security/optimizely-supply-chain-attack-delivers-malware-to-cnn-fox-news-bbc-salesforce-sites Why it matters: A suspected compromise of Optimizely led to malicious JavaScript loading on many major sites, so teams embedding third-party scripts should review those dependencies now for signs of tampering. [4] N-able Patches 17 N-central Flaws as Attackers Exploit Earlier Bugs to Deploy Crypto Miners The Hacker News · Advisory · https://thehackernews.com/2026/11/n-able-patches-17-n-central-flaws-as.html Why it matters: N-able's N-central 2026.4 Hotfix 3 is described as a critical security update, and attackers are reportedly exploiting earlier bugs to deploy crypto miners, so teams running this RMM platform should patch promptly. [5] Cloud keys leaked in Docker Hub images are stolen within seconds BleepingComputer · Research · https://bleepingcomputer.com/news/security/cloud-keys-leaked-in-docker-hub-images-are-stolen-within-seconds AGENTIC SDLC [6] The risks of device code authentication for AI coding agents Microsoft · Research · https://microsoft.com/en-us/security/blog/2026/11/17/the-risks-of-device-code-authentication-for-ai-coding-agents Why it matters: Device code phishing can be extended to AI coding agents such as GitHub Copilot CLI, so a single approved sign-in code can give attackers access; teams should review how agents authenticate. [7] External sandboxes for coding agents Simon Willison · Release · https://simonwillison.net/2026/Nov/17/external-sandboxes-for-coding-agents REMEDIATION [8] Exploited Before It Was Fixed? Vulnerability Timelines from Open, Machine-Readable Evidence arXiv · Research · https://arxiv.org/abs/2611.22479 [9] Researchers Detail iPhone Flaw That Let Apps Record Audio Despite Denied Microphone Permission The Hacker News · Research · https://thehackernews.com/2026/11/researchers-detail-iphone-flaw-that-let.html MODELS [10] SEAM: Computer-Use Agents Know the Harm Between Screens, and Commit It Anyway arXiv · Research · https://arxiv.org/abs/2611.22971 [11] Who Pays for Byzantine Robustness? Validity, Data-Share Dictatorships, and Fairness When Only a Few Hospitals Collaborate arXiv · Research · https://arxiv.org/abs/2611.22904 [12] Is 4-Bit Quantization More Than Noise to a Chat Model's Safety Behaviour? Attributing Behavioural Shifts with Matched Perturbations arXiv · Research · https://arxiv.org/abs/2611.22186 AI SECURITY [13] Network-Exposed Configuration Attacks on LLM-Controlled Social Robots arXiv · Research · https://arxiv.org/abs/2611.22761 [14] GapLeak: When Privacy-Preserving LLM-based Web Agents Become a Privacy Risk arXiv · Research · https://arxiv.org/abs/2611.22223 [15] Can You Undo It? Reversibility Classes for AI Agent Actions arXiv · Research · https://arxiv.org/abs/2611.22075 ADVISORIES IN THIS WINDOW 101 npm advisories · 99 critical · 2 high 7 PyPI advisories · 6 critical · 1 moderate 4 Packagist advisories · 1 high · 3 moderate 2 RubyGems advisories · 2 critical 1 Go advisory · 1 high 1 Maven advisory · 1 high All advisories: https://keryn.dev/headlines/advisories/ Titles, links and short excerpts only; every item links to its source. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.