🧰 GPT-5.4

The items that name it 7

• GhostSplice: Malicious MCP Servers Split Exfiltration to Bypass Guardrails
  articleas affected products2026-08-19

  Reported at 90% compliance in Cursor and 100% through Codex CLI, but dropped to 0% behind Claude Code, per ASSET Research Group's repository.

• AI Security Briefing
  articleas affected products2026-07-18

  Miscategorized the fraudulent DeBank-impersonating website as the trusted DeBank platform

• AI Security Briefing
  writeupas affected products2026-07-18
• OpenAI Codex with GPT-5.6 Sol: competitive, zero cheating, one unique Django fix | Blog | Endor Labs
  articleas tools2026-07-17

  Prior model generation tested via Codex, baseline for measuring ~8 point FuncPass and ~2 point SecPass improvement

• Endor Labs: AI Coding Agents Cheat Security Benchmarks via Training Recall, Not Reasoning
  articleas affected products2026-06-11
• Move over, Mythos. Here comes... pretty much any other model with a good harness
  articleas tools2026-06-01

  AI model; costs half as much as GPT-5.5 for inputs/outputs; used in cost-vs-performance comparisons for vulnerability discovery.

• XBOW: GPT-5.5 and Opus 4.7 in Offensive Security Workflows
  writeupas tools2026-05-06

  LLM predecessor to GPT-5.5, noted for learning to 'go faster' (fewer actions before finding vulnerabilities)

In every workspace you can read 19