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, Thursday 2 April 2026 · Retrospective · Covers 1 April 2026 00:00 to 2 April 2026 00:00 UTC

Read on the web

Top developments

Claude Code source code accidentally leaked in NPM package

BleepingComputer · Incident · Agentic SDLC, AI Security

CVE-2026-5281: Google Dawn Use-After-Free Vulnerability added to CISA's Known Exploited Vulnerabilities catalog

CISA · Advisory · Remediation

Apple expands iOS 18 updates to more iPhones to block DarkSword attacks

BleepingComputer · Analysis · Remediation

SAST

Securing the open source supply chain across GitHub

GitHub · Analysis · SAST, SCA

Obfuscating Code Vulnerabilities against Static Analysis in JavaScript Code

arXiv · Research · SAST, SCA

VibeGuard: A Security Gate Framework for AI-Generated Code

arXiv · Research · Agentic SDLC, SAST

SCA

AutoEG: Exploiting Known Third-Party Vulnerabilities in Black-Box Web Applications

arXiv · Research · DAST, SCA

DAST

Enhancing REST API Fuzzing with Access Policy Violation Checks and Injection Attacks

arXiv · Research · DAST

A Taxonomy of Cognitive Security

Schneier on Security · Analysis · DAST

OWASP PTK Findings as ZAP Alerts (Juice Shop Walkthrough)

ZAP (Checkmarx) · Analysis · DAST, SAST

Agentic SDLC

Investigating Autonomous Agent Contributions in the Wild: Activity Patterns and Code Change over Time

arXiv · Research · Agentic SDLC

Trust

Ontology-Constrained Neural Reasoning in Enterprise Agentic Systems: A Neurosymbolic Architecture for Domain-Grounded AI Agents

arXiv · Research · Trust

Introduction sigstore-c

Sigstore · Analysis · Trust

Models

SelfGrader: LLM Jailbreak Detection via Anchored Token-Level Logits

arXiv · Research · Models

Cooking Up Risks: Benchmarking and Reducing Food Safety Risks in Large Language Models

arXiv · Research · Models

ClawSafety: "Safe" LLMs, Unsafe Agents

arXiv · Research · AI Security, Models

AI Security

No Attacker Needed: Unintentional Cross-User Contamination in Shared-State LLM Agents

arXiv · Research · AI Security

AgentWatcher: A Rule-based Prompt Injection Monitor

arXiv · Research · AI Security

Evaluation and Hardening of LLM System Instructions Against Extraction via Encoding Attacks

arXiv · Research · AI Security, Models

Advisories in this window

  • 39 npm advisories · 12 critical · 15 high · 11 moderate · 1 low
  • 36 Packagist advisories · 13 critical · 8 high · 15 moderate
  • 34 PyPI advisories · 5 critical · 9 high · 14 moderate · 6 low
  • 14 Go advisories · 8 high · 3 moderate · 3 low
  • 2 Hex advisories · 1 high · 1 moderate
  • 1 NuGet advisory · 1 high

All advisories

Titles, links and short excerpts only; every item links to its source. Selection is automatic; the editor's note is AI-drafted and checked as described. Preview: the unsubscribe link and Keryn's postal address are added to each email when it is sent.