Tool Hijacking: The 2024 Papers That Predicted Agent Attacks

By November 2024, AI-agent security research had already documented the attack class that production incidents would later make infamous. InjecAgent (March 2024, ACL Findings) benchmarked 1,054 indirect-injection scenarios across 30 agents, finding ReAct-prompted GPT-4 attacked successfully roughly a quarter of the time. Breaking Agents (July 2024) demonstrated malfunction amplification through agentic loops. Together with 2023’s foundational indirect-prompt-injection work, they mapped how tools, descriptions, and fetched content become command channels. This survey walks the papers, the hijack taxonomy, and the controls that predate the incidents.

Continue ReadingTool Hijacking: The 2024 Papers That Predicted Agent Attacks

Ray AI Framework’s ‘Won’t Fix’ CVEs: A Control-Plane Debate

When Protect AI disclosed five Ray vulnerabilities in March 2024 — including critical RCE via the unauthenticated control plane — Anyscale’s ‘won’t fix, trusted-networks design’ stance ignited the year’s sharpest debate over AI infrastructure responsibility. This piece unpacks the job-submission RCE, the exposed-cluster census, the bounty economics, what Anyscale later shipped anyway, and the hardening playbook that became standard for every exposed ML control plane.

Continue ReadingRay AI Framework’s ‘Won’t Fix’ CVEs: A Control-Plane Debate
Read more about the article Prompt Injection Is the New SQL Injection: The 20-Year-Old Mistake AI Is Repeating in 2026
Prompt injection is the new SQL injection — featured image

Prompt Injection Is the New SQL Injection: The 20-Year-Old Mistake AI Is Repeating in 2026

SQL injection stayed in the OWASP Top 10 for 20+ years. Prompt injection is the same bug with worse permissions — here’s the history, real examples, and the defense playbook.

Continue ReadingPrompt Injection Is the New SQL Injection: The 20-Year-Old Mistake AI Is Repeating in 2026