Gray Swan AI

§ Positioning

Runs adversarial arenas and red-teaming programs that probe models and agents for failures, feeding safety evaluations back to developers.

Best for · Safety and robustness evaluation under adversarial pressure.

Team · Founding team from academic AI-safety research (verify before publishing).

Focus · Red-teaming and safety-evaluation arenas that adversarially stress-test models and agents.

Notable customers · Frontier labs (per company)

• Long Horizon · Evaluation

  Andon Labs

  agent benchmarks with a sense of humor

• Data Curation · Evaluation

  Bespoke Labs

  open data and reproducible recipes

• Long Horizon · Code

  Huzzle Labs

  environments + human data + evals in one stack

← Browse all companies