TPRM Benchmark · SecurityScorecard
Join the founding research cohort
Help validate, run, and expand the benchmark before public launch.
For academia, model labs, enterprises, public-sector teams, and TPRM practitioners.
Public before sign-in
Read the benchmark thesis, methodology, aggregate release coverage, and representative open samples.
Active participation through Driftnet
A free verified account opens the private workspace. Connect your provider or custom endpoint, run one model or a cohort, inspect private evidence, propose tasks or tracks, and request the versioned local research package.
Bring your own model
The workspace supports major hosted providers, AWS Bedrock, editable OpenAI-compatible connections, and custom HTTPS endpoints for open-weight or self-hosted models. Provider model catalogs populate a dropdown when the account permits discovery.
Bring your own key
Benchmark prompts and outputs are sent to the provider you select and may appear in its logs. Its retention, usage, and billing terms apply. Credentials are encrypted, scoped to your account, and never returned by the service.
Research governance
Your evaluations remain private unless submitted for review. Official cohort evidence requires completion, reproducible configuration checks, and designated reviewer approval. Accepted proposals still require implementation and reference-solver admission before release.