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 30 September 2026 Retrospective Covers 29 September 2026 00:00 to 30 September 2026 00:00 UTC Web: https://keryn.dev/headlines/daily/2026-09-30/ TOP DEVELOPMENTS [1] Citrix patches actively exploited NetScaler zero-days after a weekend of unofficial warnings CyberScoop · Advisory · https://cyberscoop.com/citrix-zero-days-delayed-disclosure [2] CVE-2026-86950: Apple Multiple Products Out-of-Bounds Write Vulnerability added to CISA's Known Exploited Vulnerabilities catalog CISA · Advisory · https://cisa.gov/known-exploited-vulnerabilities-catalog?search_api_fulltext=CVE-2026-86950 [3] Kiteworks lifts shutdown advisory after ‘credible threat intelligence’ from federal authorities CyberScoop · Advisory · https://cyberscoop.com/kiteworks-lifts-shutdown-advisory-after-credible-threat-intelligence-from-federal-authorities [4] Automated AI agent used to breach cybersecurity nonprofit DIVD BleepingComputer · Incident · https://bleepingcomputer.com/news/security/automated-ai-agent-used-to-breach-cybersecurity-nonprofit-divd [5] Kiteworks patches critical flaw, brings customer systems online BleepingComputer · Advisory · https://bleepingcomputer.com/news/security/kiteworks-lifts-shutdown-warning-after-patching-critical-flaw SAST [6] Retrieve, Reproduce, Reveal: Dissecting Retrieval-Augmented Software Vulnerability Detection arXiv · Research · https://arxiv.org/abs/2609.37669 SCA [7] 101 Malicious npm Packages Add Developers' WhatsApp Accounts to Groups Without Consent The Hacker News · Research · https://thehackernews.com/2026/09/101-malicious-npm-packages-add.html [8] Concealing LLM-Based Multi-Agent Topology via Phantom Structure Injection arXiv · Research · https://arxiv.org/abs/2609.37567 [9] SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs arXiv · Research · https://arxiv.org/abs/2609.36879 AGENTIC SDLC [10] SecureVibe: Making Vibe Coding More Secure arXiv · Research · https://arxiv.org/abs/2609.38606 [11] HARDE: Optimizing Agent Harnesses for Runtime Risk Detection and Execution Control arXiv · Research · https://arxiv.org/abs/2609.38291 [12] Practical Secrets Extraction against Black-box LLMs arXiv · Research · https://arxiv.org/abs/2609.36941 REMEDIATION [13] One Pipeline Does Not Fit All: TAILOR, a Type- and State-Aware Framework for CVE Reproduction arXiv · Research · https://arxiv.org/abs/2609.37006 [14] WitnessGym: Benchmarking Coding Agents on the Construction of Bug Witnesses arXiv · Research · https://arxiv.org/abs/2609.36635 TRUST [15] CRA Tech Talk, 09/07/2026 OpenSSF · Standard / policy · https://openssf.org/policy/cra/2026/09/29/cra-tech-talk-09-07-2026 [16] Fault-Tolerant Budget Conservation in Distributed Multi-Agent Delegation arXiv · Research · https://arxiv.org/abs/2610.00349 [17] AutoMark: Enabling Autoresearch to Discover Better LLM Watermarks arXiv · Research · https://arxiv.org/abs/2609.37310 MODELS [18] Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation arXiv · Research · https://arxiv.org/abs/2610.00348 [19] A Competing-Hazards Systematization of Loss of Control in Autonomous Agents arXiv · Research · https://arxiv.org/abs/2609.38411 [20] Aegis: Generative Gradient Masking for Privacy-Preserving Medical Federated Learning arXiv · Research · https://arxiv.org/abs/2609.38339 AI SECURITY [21] Where Do LLMs Decide to Break the Rules? Mechanistic Localization of Prompt Injection Compliance arXiv · Research · https://arxiv.org/abs/2609.37737 [22] ToolFence: Fine-Grained Authorization for Secure Tool-Using LLM Agents arXiv · Research · https://arxiv.org/abs/2609.37196 [23] Selecting The Most Informative Tokens in Natural Language Autoencoders arXiv · Research · https://arxiv.org/abs/2609.37040 ADVISORIES IN THIS WINDOW 94 npm advisories · 57 critical · 17 high · 17 moderate · 3 low 13 PyPI advisories · 3 critical · 5 high · 5 moderate 3 Packagist advisories · 1 high · 2 low 2 Go advisories · 2 moderate 1 crates.io advisory · 1 moderate 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.