# Sometimes ease and security are totally inverse. **By:** Ken C. **Published:** 2026-05-16T11:48:17.924Z **Source:** [LinkedIn](https://www.linkedin.com/posts/ken-corigliano_ai-cio-cto-share-7461382013271367680-jMBr) --- # Ken C.’s Post [![View profile for Ken C.](https://static.licdn.com/aero-v1/sc/h/9c8pery4andzj6ohjkjp54ma2)](https://www.linkedin.com/in/ken-corigliano?trk=public_post_feed-actor-image) [Ken C.](https://www.linkedin.com/in/ken-corigliano?trk=public_post_feed-actor-name) Sometimes ease and security are totally inverse. As an [#AI](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fai&trk=public_post-text) expert, my primary job is not to push or flood everywhere and everything with AI. It is to protect status quo from it while maximizing its potential. My colleagues are not doing this. If you have a [#CIO](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fcio&trk=public_post-text), [#CTO](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fcto&trk=public_post-text), [#CDAO](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fcdao&trk=public_post-text), [#CAIA](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fcaia&trk=public_post-text), etc they need to be pushing back on shiny objects as much as entertaining them. This is exactly why I refused to employ [#Claude](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fclaude&trk=public_post-text). Anyway, here's the plain language: AI didn't just play a background role in the Mini [#Shai](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fshai&trk=public_post-text)\-Hulud campaign. It was a core pillar of the attack strategy. TeamPCP intentionally weaponized the tech industry's massive shift toward artificial intelligence, using AI both as a high-value target and as a camouflage mechanism to bypass human suspicion. AI intersected with this supply-chain worm in three major ways: 1. Camouflage via AI Agent Impersonation Because developers are now highly accustomed to seeing automated AI coding assistants and bots making routine dependency updates in their repositories, these malicious commits went unnoticed by maintainers for critical windows of time. 2. Highjacking Local AI Coding Assistants for Persistence Specifically, it looked for Claude Code (Anthropic’s command-line AI agent) and VS Code installations on the developer's machine. 3. Aggressively Targeting the AI Supply Chain The malware was explicitly coded to hunt for environment variables, config files, and API keys belonging to [#LLM](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fllm&trk=public_post-text) providers. Stolen AI API keys and model access tokens are currently incredibly lucrative on the cybercriminal black market. Significance: In this campaign, the [#hackers](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fhackers&trk=public_post-text) impersonated the AI tools that developers have grown to trust implicitly, proving that automated workflows are now prime targets for exploitation. This attack marks a paradigm shift. Historically, hackers impersonated human colleagues via phishing. [Anthropic](https://www.linkedin.com/company/anthropicresearch?trk=public_post-text) [#hacking](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fhacking&trk=public_post-text) [#cybersecurity](https://www.linkedin.com/signup/cold-join?session_redirect=https%3A%2F%2Fwww.linkedin.com%2Ffeed%2Fhashtag%2Fcybersecurity&trk=public_post-text) [Alex Turnbull](https://www.linkedin.com/in/alex-turnbull-1ab9992?trk=public_post_reshare_feed-actor-name) WARNING: The Shai-Hulud npm exploit has INFECTED 518M computers. It bypasses 2FA, hides in Claude Code, & nukes your home dir. if you revoke the GitHub token it stole. Here's how to triage ASAP 👇 TeamPCP forked TanStack's repo and tricked its release pipeline into signing 84 malicious versions across 42 official packages. @tanstack/react-router alone get 12M+ weekly downloads. To npm, to your CI and to every SLSA verifier on earth the packages looked 100% legitimate. The entire TanStack team had 2FA on and it did not matter. It's already crossed to PyPI in 48 hours. Confirmed hist: @tanstack/react-router @opensearch-project/opensearch mistralai (PyPI) guardrails-ai (PyPI) @uipath/apollo-core @squawk/\*, SAP packages And OpenAI corporate. OpenAI is rotating code-signing certs and force-updating every macOS user. You don't rotate signing certs for "limited credential material." You rotate them when the attacker was one step from signing binaries as you. CVE-2026-45321. CVSS 9.6. is active in the wild. The malware steals every credential on your machine, hooks into ~/.claude/[settings.json](http://settings.json?trk=public_post_reshare-text) and ~/.vscode/[tasks.json](http://tasks.json?trk=public_post_reshare-text) to re-execute after uninstall, and plants a watcher that triggers rm -rf ~/ the moment you revoke its GitHub token. What to triage RIGHT NOW in this exact order: 1 / Check before touching GitHub: grep lockfiles for affected packages after May 11. Check ~/.claude/[settings.json](http://settings.json?trk=public_post_reshare-text), ~/.vscode/[tasks.json](http://tasks.json?trk=public_post_reshare-text), and ps aux | grep gh-token-monitor. 2 / Defuse before rotating: pkill -f gh-token-monitor. Remove IDE hooks. Block [git-tanstack.com](http://git-tanstack.com?trk=public_post_reshare-text). 3 / Rotate everything: GitHub, AWS/GCP/Azure, npm, PyPI, SSH, CI secrets, browser passwords, signing certs. 4 / Reinstall: wipe node\_modules, pin to pre-May 11 versions, audit packages you maintain. R u at SoF?