Fabraix
Fabraix builds frontier red-teaming AI agents that continuously find security vulnerabilities in customer-facing AI systems.
At a Glance
- AI Developer Teams
- Fortune 500 Enterprises
- Cybersecurity Departments
AI Tools by Fabraix
(1)
Fabraix Playground
AI Agent Security Challenge Platform
Discussions
No discussions yet
Be the first to start a discussion about Fabraix
Latest News
Bounding the Blast Radius: A Survey of Prompt-Injection Defenses for LLM Agents
Adversarial Cost to Exploit (ACE): A Dynamic Benchmark for AI Agent Security
2026 so far has matched all of 2024 for AI security incidents
SWE-bench will hit 90% this year
Products & Services
An autonomous testing harness for AI agents that probes security, logic, and alignment using adaptive, multi-turn blackbox attacks.
A dynamic benchmark measuring the token expenditure required for an adversary to breach an LLM-backed agent.
An open-source adversarial testing environment for running attack campaigns against AI systems.
Market Position
Automates the work of manual pentest teams by providing continuous, adaptive, and autonomous red-teaming at scale.
Leadership
Founders
Ahmed Aly
Previously the first data scientist at Sequoia-backed Two, where he built a fraud engine processing $1B+ in annual transactions. Published researcher with papers in Q1 journals and a UCL PhD dropout.
Ibrahim Abdu
Previously a Software Engineer at Meta building AI agents for production errors. Early engineer at Two and built a proprietary compiler/database at TradingHub. Oxford graduate (PPE).
Executive Team
Ahmed Aly
Co-founder
Expert in data science and fraud prevention; former lead at Sequoia-backed Two.
Ibrahim Abdu
Co-founder
Former Meta SWE specializing in AI agents and TradingHub engineer.
Board of Directors
Founding Story
Founded by security researchers from Meta and Oxford to automate the expensive and slow manual penetration testing process for non-deterministic AI agents.
Business Model
Revenue Model
Subscription-based for continuous monitoring and usage-based/custom fees for one-off audits and enterprise deployments.
Pricing Tiers
Free access for academic and independent AI safety research via application.
Full adversarial audit of a single agent with AIVSS-scored pentest report.
Monthly usage-based pricing for CI/CD integration and scheduled scans.
Dedicated infrastructure, SSO, and compliance coverage (EU AI Act, ISO 42001).
Target Markets
- AI Developer Teams
- Fortune 500 Enterprises
- Cybersecurity Departments
- Customer support bots
- Coding agents
- Financial advisors
- Clinical copilots
- Voice assistants
- RL reward hacking detection
- Dozens of Fortune 500 companies