🧰 DeepSeek V4 Pro

The items that name it 7

• Structured Security Explanations for Learning-Based Vulnerability Detection
  paperas tools2026-10-27

  LLM baseline for vulnerability judgment; VulnLens exceeds it by 4.2 points in accuracy on PrimeVul, and by 7.6 points in accuracy and 20.0 points in F1 on SecVulEval.

• ‘Context Bombs’ Cut AI Attack Success by Misleading Agents With Fake Targets
  articleas affected products2026-10-21It is written here as DeepSeek V4 Pro: attacks stopped by all nine context bombs.
• US Orders Civilian Agencies to Rip DeepSeek Out of Government Systems
  articleas affected products2026-10-19It is written here as DeepSeek V4-Pro (preview released April 2026, MIT license).

  preview released April 2026, MIT license

• ZeroShield - Securing AI Infrastructure (@zeroshieldai) on X
  threadas tools2026-10-01

  Open-weight AI model benchmarked finding CVEs in Aikido's test; found 28 of 32 recent CVEs for ~$98 a run

• Autonomous AI Agents Hack Online Retailers for $25 a Company
  writeupas tools2026-09-22

  AI model later used to run Strix scans

• [tl;dr sec] #345 - Bug Rumors → Exploits, Version Control DFIR, Agentic Worms
  articleas tools2026-09-10

  Model that wrote an exploit for the cohttp path traversal bug in under a minute from a vague description.

• Rumour Is the Exploit — AI agents collapse time-to-exploit to near-zero
  articleas tools2026-08-22It is written here as DeepSeek V4 Pro (open-weight model used by the author to find several related path normalisation issues).

  open-weight model used by the author to find several related path normalisation issues

In every workspace you can read 24