Solayer Founder Reveals LLM Supply Chain Major Security Flaw: Over 20% Free Routes Exposed to Malicious Injection

On April 10, Solayer founder @Fried_rice took to social media to note that Large Language Model (LLM) agents are increasingly relying on third-party API routers—tools that route tool invocation requests to multiple upstream providers. These routers act as application-layer proxies, capable of accessing every payload in plaintext during transit. Critically, no provider currently enforces end-to-end encryption and integrity protection between clients and upstream models. A research paper tested 28 paid routers (purchased from Taobao, Xianyu, and Shopify standalone shops) and 400 free routers (collected from public communities). Key findings: - 1 paid router and 8 free routers actively inject malicious code; - 2 deploy adaptive evasion triggers; - 17 accessed AWS Canary credentials belonging to researchers; - 1 stole ETH from a researcher-held private key. Two poisoning studies further highlight risks from seemingly benign routers: - A leaked OpenAI key generated 1 billion GPT-5.4 tokens and over 7 Codex sessions; - Low-configured bait led to 20 billion billing tokens, 99 credentials across 440 Codex sessions, and 401 sessions running in autonomous YOLO mode. The research team built a research-focused agent called Mine, which can execute all four attack types against four public proxy frameworks. They also validated three client-side defense measures: fault-locking policy gating, response-side anomaly screening, and append-only transparent log recording.