Why this exists: on Oct 1, 2026 the FTC opened an industry-wide probe into Anthropic, OpenAI, and METR over rogue AI agents — the first US enforcement action on agent behavior (Reuters). Gravitee's State of AI Agent Security 2026: 88% of orgs confirmed or suspected an agent security incident in the past year (750 execs surveyed). Probe your own agent before the FTC does.
① Pick an attack
② Toggle defenses
③ Fire the probe
🛡️ blocked 0
💥 breached 0
attempts: 0
each fire = 1 metered attempt (try the API for the real 402)
probe-lab v0.1.0 — select an attack, set your defenses, then fire.hint: start with defenses OFF. watch everything breach. then turn them ON one at a time.
Red-team your agent like you mean it
This demo is the free teaser. The ProbeLab kit ($39, one-time) ships the full red-team package: the 12-technique attack catalog with example payloads, the 7-defense checklist mapped attack-by-attack, a drop-in FastAPI guardrail middleware (per-key attempt budgets, hard 402 cutoff, kill switch), and a scoring rubric to grade your agent before the next incident report grades you.
$79 → $39 one-time · yours forever
No subscription. No tokens. Educational payloads only — nothing weaponized.