Alice
alice.ioSecurity
Hall of Fame—
Known for
Finding out how your own model can be talked into misbehaving before a customer finds out for you.
Red-teams an AI application before launch, then keeps guardrails on it in production and re-tests it as the model drifts. Runs on a corpus of adversarial samples rather than a fixed rule list.
What it does6
- Pre-launch stress testing of models, apps and agents
- Runtime guardrails on live applications
- Continuous automated red-teaming and drift detection
- Adversarial sample corpus behind real-time threat detection
- Safety policy tuned per regulation and risk tolerance
- Covers text and images across 120+ languages
From Alice’s own site. Not used for ranking.
Too big to rank
Past $140M raised or earned. Too big to rank.
Money$140M · above the board
Compared with everyone else on this board.