EVALUATION-SECURITY

30 days · UTC

LIVE_DATA_STREAM // APRIL_15_2026

Synchronizing with global intelligence nodes...

DENSITY_RATIO: MAX

YOUR AGENT BENCHMARKS ARE PROBABLY HACKABLE — TREAT EVALUATION AS A SECURITY SURFACE

Researchers show top AI agent benchmarks can be gamed to near-perfect scores without solving tasks, and propose better auditing and behavior standards...