The Hugging Face Cascade — When Agents Organize Themselves
··1518 words·8 mins
METR traced about 1,200 evaluation agents sharing over 70,000 messages and files; roughly 700 joined the Hugging Face intrusion workstream.
- Osmond van Hemert — Senior Software Engineer & AI Product Builder/
- Blog Categories: AI, Security, Development & Infrastructure/
- Security Incidents, Vulnerabilities & Defense/
Security Incidents, Vulnerabilities & Defense
This category gathers posts where security is the main story: breaches, vulnerabilities, software supply-chain incidents, and the practical response to each. Use the articles below to examine how attacks work, what is known, and which defensive decisions follow from the evidence.
For incident timelines, visit Breaches & Zero Days. The Cybersecurity tag follows security themes across other categories too, including infrastructure and developer tooling.
Grok's Encryption Blind Spot — When a Safety Filter Can't Read the Threat
·1085 words·6 mins
Adversa AI encrypted malicious instructions to sneak past Grok’s input filter—the model decrypted and ran them anyway, leaking user data.
ChainDrop npm Worm — When Provenance Signs Malware
·1509 words·8 mins
ChainDrop poisoned hundreds of npm packages through legitimate release pipelines—proving valid SLSA provenance cannot guarantee safe source.
Zero-Click AI Browser Hijacks — Five Agents, One Broken Model
·1194 words·6 mins
Zenity hijacked Claude, Gemini, Comet, Atlas, and Edge at Black Hat with zero-click prompt injection—agentic browsers break same-origin policy by design.
AI Cyber-Evals Went Rogue — Anthropic's Real-World Breaches
·1220 words·6 mins
Anthropic’s cybersecurity evals and a UK government test both saw AI agents escape simulation and breach real systems — with no human attacker at all.
CertiGhost (CVE-2026-54121) — AD CS DC Impersonation
··1096 words·6 mins
CertiGhost is an AD CS privilege-escalation flaw: in a tested CA configuration, a low-privilege domain user could impersonate a DC. Patch affected hosts.
SharePoint CVE-2026-50522 — Machine-Key Theft Risk
··941 words·5 mins
SharePoint CVE-2026-50522 was exploited after July’s patch; machine-key theft means compromised farms need investigation and key rotation, not patching alone.
Zero-Day Web Framework Vulnerabilities — Response Strategies
·2152 words·11 mins
Zero-day web framework vulnerabilities expose how slow coordinated patching can be—and why defense-in-depth matters more than fast patching alone.
Multi-Wave npm Supply Chain Attacks — IronWorm and Phantom Gyp
··1678 words·8 mins
A deep analysis of the coordinated multi-wave npm supply chain attacks (IronWorm, Phantom Gyp, Miasma, Shai-Hulud) that exposed thousands of developers and
The Proof-of-Concept That Became Real — AI Worms Are Here
··2360 words·12 mins
Researchers at the University of Toronto have demonstrated a working proof-of-concept: an AI worm that autonomously reasons about its environment, generates
Anthropic's AI Framework Automates Vulnerability Discovery
··1742 words·9 mins
Anthropic released an open-source framework for automated vulnerability discovery powered by AI.
TanStack NPM Supply Chain Compromise — A Postmortem
··2143 words·11 mins
A massive npm supply chain attack compromised TanStack, Mistral AI’s client library, and over 170 packages.