Where intelligence meets accountability.
The independent record of how far autonomous AI can be trusted.
TRUSCOR builds no models, no agents and no applications. We test what AI systems do, record it as signed evidence, and rate how far they can be trusted.
CIBIL scores borrowers and never lends a rupee, which is why every bank trusts it. We are the bureau, not the bank.
| What it does | Access | |
|---|---|---|
| SOVA OSS | Free toolkit to map, test, record, replay and investigate AI agents | Open source · Apache-2.0 |
| .sova format | A signed, tamper-evident record of what an AI was shown, what it did and what changed | Open |
| SOVA Engine | Attacks, UI-tests and stress-tests AI products and writes .sova evidence |
Rating team and licensed security firms |
| SOVA Black Box | A 24x7 recorder for AI in production, signed and checked against effect | Production deployments and OEM partners |
| O.W.L | The measurement instrument: checks what an AI claimed against what really happened | SOVA Engine, Black Box and licensed measurers |
| TAFAAR | Grades evidence into the TRUSCOR Rating (300–900) and an Estimated Financial Exposure | TRUSCOR only, never licensed |
No model, API key or network needed:
git clone https://github.com/TRUSCOR/SOVA-OSS
cd SOVA-OSS
uv sync --locked
uv run sova demo sleeperTRUSCOR/research holds our released papers and R&D with their data,
code, protocols and evidence as .sova files.
The latest preprint, .sova: Portable, Evidence-Bounded Reproduction of AI-Agent Security Findings (version 1.1, November 2026, not yet peer reviewed), re-ran a widely used AI-agent security benchmark on eight models and ships four findings you can verify offline: try one.
- No product in any category we rate
- No referral economics
- No certification today
- No insurance on our balance sheet
SOVA OSS is open to contributions: start with the issues at TRUSCOR/SOVA-OSS.
TRUSCOR was founded by Gautam R. Patil, who created SOVA and the .sova format.
