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, Saturday 14 March 2026 Retrospective Covers 13 March 2026 00:00 to 14 March 2026 00:00 UTC Web: https://keryn.dev/headlines/daily/2026-03-14/ TOP DEVELOPMENTS [1] Google fixes two new Chrome zero-days exploited in attacks BleepingComputer · Incident · https://bleepingcomputer.com/news/google/google-fixes-two-new-chrome-zero-days-exploited-in-attacks [2] CVE-2026-3909: Google Skia 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-2026-3909 [3] Supply-chain attack using invisible code hits GitHub and other repositories Ars Technica: Security · Incident · https://arstechnica.com/security/2026/03/supply-chain-attack-using-invisible-code-hits-github-and-other-repositories AGENTIC SDLC [4] Human-Centered Evaluation of an LLM-Based Process Modeling Copilot: A Mixed-Methods Study with Domain Experts arXiv · Research · https://arxiv.org/abs/2603.12895 [5] SWE-Milestone: Evaluating AI Agents on Continuous Software Evolution arXiv · Research · https://arxiv.org/abs/2603.13428 REMEDIATION [6] Microsoft: Windows 11 users can't access C: drive on some Samsung PCs BleepingComputer · Analysis · https://bleepingcomputer.com/news/microsoft/microsoft-windows-11-users-cant-access-c-drive-on-some-samsung-pcs [7] Defensible Design for OpenClaw: Securing Autonomous Tool-Invoking Agents arXiv · Research · https://arxiv.org/abs/2603.13151 TRUST [8] Colluding LoRA: A Compositional Vulnerability in LLM Safety Alignment arXiv · Research · https://arxiv.org/abs/2603.12681 MODELS [9] Purifying Generative LLMs from Backdoors without Prior Knowledge or Clean Reference arXiv · Research · https://arxiv.org/abs/2603.13461 [10] Test-Time Attention Purification for Backdoored Large Vision Language Models arXiv · Research · https://arxiv.org/abs/2603.12989 AI SECURITY [11] PISmith: Reinforcement Learning-based Red Teaming for Prompt Injection Defenses arXiv · Research · https://arxiv.org/abs/2603.13026 [12] Uncovering Security Threats and Architecting Defenses in Autonomous Agents: A Case Study of OpenClaw arXiv · Research · https://arxiv.org/abs/2603.12644 [13] Agent Privilege Separation in OpenClaw: A Structural Defense Against Prompt Injection arXiv · Research · https://arxiv.org/abs/2603.13424 ADVISORIES IN THIS WINDOW 72 npm advisories · 36 critical · 19 high · 16 moderate · 1 low 11 Go advisories · 3 critical · 1 high · 6 moderate · 1 low 7 PyPI advisories · 5 high · 2 moderate 5 crates.io advisories · 4 high · 1 moderate 3 Packagist advisories · 2 high · 1 moderate 2 Maven advisories · 2 moderate 2 NuGet advisories · 2 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.