Nyx
Autonomous, pure-blackbox adversarial testing harness that probes AI agents across security, logic, and alignment using multi-turn, adaptive, multi-modal stress tests.
What it does
Autonomous, pure-blackbox adversarial testing harness that probes AI agents across security, logic, and alignment using multi-turn, adaptive, multi-modal stress tests.
- AI red-teaming
- adversarial verification
- security vulnerability discovery
- Category
- AI safety and governance
- Users
- Fortune 500 companies, enterprise AI teams, academic and independent researchers
- Public API
- Not offered
Not yet disclosed: deployment options, platforms, languages and support channels.
Product insights
Evidence for this product in each industry and task.
Documented feature
“path to safe superintelligence runs through offensive AI. Every failure surfaced is intelligence we didn't have yesterday. A new failure mode found leads to a new guardrail trained. The faster we break AI, the faster AI gets…”
Read product documentation
path to safe superintelligence runs through offensive AI. Every failure surfaced is intelligence we didn't have yesterday. A new failure mode found leads to a new guardrail trained. The faster we break AI, the faster AI gets…
Read source · fabraix.com ↗path to safe superintelligence runs through offensive AI. Every failure surfaced is intelligence we didn't have yesterday. A new failure mode found leads to a new guardrail trained. The faster we break AI, the faster AI gets…
Read source · fabraix.com ↗
Pricing
- Free plan$0
- Free access for academic
- Independent non-commercial research
- By application
- EnterpriseCustom
- Quoted by their sales team
Prices as published by Fabraix, Inc., checked Nov 2026.
Security and compliance
As reported by Fabraix, Inc.. Ask for audit reports before you buy.
- SOC 2 Type IIIn progress
- EU AI ActCompliant
- ISO/IEC 42001Compliant
- MITRE ATLASCompliant
- TLS 1.2+ encryption in transit
- AES-256 encryption at rest
- Isolated, ephemeral sandboxed scan environments
- Synthetic-only payloads with canary secrets
- Pure blackbox testing with no internal access
- Never trains on customer data
- Least-privilege, logged access
- Responsible-disclosure program
- Minimal, configurable data retention with deletion on request
- Privacy-respecting analytics (Umami, no cross-site tracking)
Alternatives
See the full rankingNVIDIA NeMo
by NVIDIA
#1 in Evaluate AI systems
Enterprise
by Mercor
#2 in Evaluate AI systems
AI App Testing
by TesterArmy
#3 in Evaluate AI systems
Opslane
by Opslane, Inc.
#5 in Evaluate AI systems
Tau2-Infinity
by Vibrant Labs
#6 in Evaluate AI systems
#7 in Evaluate AI systems
Is this your product?
Claim it to keep details accurate and submit evidence for the next ranking review.
Claim Nyx