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, Friday 10 October 2025 · Retrospective · Covers 9 October 2025 00:00 to 10 October 2025 00:00 UTC

Read on the web

Top developments

Pattern Enhanced Multi-Turn Jailbreaking: Exploiting Structural Vulnerabilities in Large Language Models

arXiv · Research · Models

CommandSans: Securing AI Agents with Surgical Precision Prompt Sanitization

arXiv · Research · AI Security

Building Security in Open Source for Financial Services: OpenSSF at Open Source in Finance Forum (OSFF)

OpenSSF · Analysis · SCA, Trust

DAST

Bug Histories as Sources of Compiler Fuzzing Mutators

arXiv · Research · DAST

Agentic SDLC

From Assistant to Adversary: Exploiting Agentic AI Developer Tools

NVIDIA · Analysis · Agentic SDLC

Models

VisualDAN: Exposing Vulnerabilities in VLMs with Visual-Driven DAN Commands

arXiv · Research · Models

Backdoor Vectors: a Task Arithmetic View on Backdoor Attacks and Defenses

arXiv · Research · Models

MetaDefense: Defending Finetuning-based Jailbreak Attack Before and During Generation

arXiv · Research · Models

AI Security

Invisible to Humans, Triggered by Agents: Stealthy Jailbreak Attacks on Mobile Vision-Language Agents

arXiv · Research · AI Security, Models

When AI Remembers Too Much – Persistent Behaviors in Agents’ Memory

Palo Alto Networks · Analysis · AI Security

Advisories in this window

  • 196 npm advisories · 194 critical · 2 high
  • 6 PyPI advisories · 3 critical · 1 high · 2 moderate
  • 2 Maven advisories · 2 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.