Fabraix

A security platform that uses autonomous AI agents to continuously probe customer-facing AI systems for vulnerabilities, addressing the problem of non-deterministic AI agents that break in unpredictable ways. It sells to security teams and executives at enterprises, particularly those in finance, healthcare, retail, and technology. The platform is positioned as state-of-the-art, with a self-improving attack engine that outperforms frontier models on benchmarks, and is delivered as a SaaS with a free tier.

Key features

• State-of-the-art attack success rate
• Adaptive attacks with 10,000+ strategies
• Self-improving attack loop
• Probes direct and indirect vulnerabilities
• Covers chatbots, agents, multi-agent systems
• Industry-specific compliance probes
• Case studies of real breaks