🔬 Research & Data · Free

VERITAS Omega Agent Trust Lab

Blindly test whether agent action packets are safe or forged before revealing the answer.

69
Views
0
Saves

About VERITAS Omega Agent Trust Lab

A free, open-source browser lab with six blind cases covering forged verdicts, parameter swaps, nonce replay, correlated reviewers, deleted evidence, and stale monitoring. Users seal ALLOW/BLOCK labels before reveal and can download a deterministic result packet; it never authorizes execution.

Updated 7/29/2026

Key Features

  • Blind calibration — Six test cases covering forged verdicts, parameter swaps, nonce replay, correlated quorum, evidence deletion, and silent monitor.
  • Zero signup — Runs locally in the browser with no account required.
  • Deterministic results — Seal your decisions before reveal and download a tamper-proof result packet.
  • No execution authority — The tool never authorizes execution; it's strictly for verification.
  • Interactive challenges — Six interactive cases where you choose ALLOW or BLOCK before seeing the answer.
  • Mechanism inspection — After each challenge, inspect how the verdict was reached and why.
  • Attack surface awareness — Learn about six common attack vectors against agent verification systems.
  • Open source — Free and open source for community scrutiny and improvement.

Pricing Structure

  • Free — $0 — Free open-source tool with no paid tiers.

See current plans at VERITAS Omega Agent Trust Lab.

Why People Use It

Teams use VERITAS to blind-test agent verdicts before revealing answers, ensuring evaluations aren't influenced by seeing the output first. The platform teaches six common attack vectors that can compromise agent verification systems, helping builders harden their AI safety practices. By sealing decisions and providing deterministic result packets, it enables reproducible, trustworthy assessments of agent behavior.

Reviews

No reviews yet. Be the first!