Grey-market relays reroute Codex and Claude Code traffic
Symantec says lookalike-domain scripts disable sandboxes and send prompts, files and tokens through operator servers at about a dozen firms.
13
Stories per day · 30 days
2 RSS
Relatedclaude-code3hardcoded-keys2npm2prompt-injection2reverse-engineering2usb-encryption2agents-md1binary-ninja1coding-agents1credential-theft1grey-market1key-management1malfex1oauth1sandbox1
10 stories, newest first · one card per story
Stories tagged #openai-codex
Researcher uses Codex to bypass four USB encryption products
Hidden accounts and hardcoded keys decrypted protected files without owner passwords, one flaw in seven minutes.
2
Planted Downloads files can hijack Claude Code and Codex agents
Forever Security says a planted AGENTS.md made Codex install MALFEX package function-color in over 90 percent of tests.
1
OpenAI Codex recovers keys from four encrypted USB products
Hardcoded AES, SM4, and Triple-DES keys and hidden accounts let files be decrypted without the user password.
3
Malicious AGENTS.md files in Downloads hijack Codex and Claude Code
A customer agent tried to install npm package function-color; tests also stole Claude Code OAuth tokens in about half of runs.
2
Decepticon – Autonomous Red Team Agent that Integrates Claude and Codex
Open-source Decepticon red team agent links Claude Code and Codex to a LangGraph runtime for scoped attack-path testing.
14
Pwn2Own Hackers Find 32 Zero-Day Vulnerabilities on Day One
Day one of Pwn2Own Ireland 2026 yielded 32 demonstrated zero-days against phones, smart home devices, printers, and AI tools, earning $368,000.
Unity launches official plugins for Claude Code and OpenAI Codex to stop AI agents from using outdated tutorials
Unity released official plugins for Claude Code and OpenAI Codex, supplying maintained skills so AI agents stop using outdated engine tutorials.
Context Engineering Inside the Harness: 4 Mechanisms That Beat Context Overflow and Goal Loss on Long-Horizon Tasks
Survey of four harness mechanisms—context budgeting, compaction, todo-state, and memory—that keep long-horizon LLM agents on task across 200+ tool calls.
9
AI Coding Agents Are Installing Unknown/Untrusted Code on Corporate Networks
Researchers found 120 corporate llms.txt files pointing to unregistered packages, demonstrating AI coding agents install and execute attacker-controlled code on Fortune 500 networks.
2