Fabraix

The world's frontier hacker of AI agents.

3.3Score#277 of 999 agents#6 of 12 in Pentester

Why it ranks #277

Fabraix, an AI pentester from YC S26, ranks #277 of 999 AI agents and #6 of 12 in Pentester, with a score of 3.3 of 10. Data as of 3 Oct 2026.

The score adds four parts, each 0 to 10, by weight. Every point comes from public evidence. How scoring works ·

Proof 2 × 30% 0.6
Scale 1 × 30% 0.3
Momentum 6 × 25% 1.5
Autonomy 6 × 15% 0.9
= 3.3

Proof
Do customers use it? 30% of the score
2
Claims a customer count.
- +2
  A customer count
  “Our product, Nyx, has already found vulnerabilities in agents at dozens of Fortune 500 companies, and hits a 78% attack success rate on AgentHarm (leading offensive AI security benchmark) vs.”
  Launch YC post · 20 Jul 2026 · 3 claims
Not found: revenue or growth (+4), named customers (+3), a customer result (+2), work done at scale (+2)
Scale
Is there a real company behind it? 30% of the score
1
2 people.
- +1
  2 people on its YC page2 people
Not found: funding round in its news (+3)
Momentum
Is it shipping and growing now? 25% of the score
6
Shows a recent launch, a new claim and a new batch.
- +3
  Launched in the last 6 monthsFabraix: The Frontier Agent that Hacks Customer-facing AI.Launch YC post · 20 Jul 2026
- +2
  A new traction claim
  “Our product, Nyx, has already found vulnerabilities in agents at dozens of Fortune 500 companies, and hits a 78% attack success rate on AgentHarm (leading offensive AI security benchmark) vs.”
  Launch YC post · 20 Jul 2026
- +1
  New: YC Summer 2026
Not found: open roles on YC (+2), funding news in the last year (+2)
Autonomy
How much of the job does it do on its own? 15% of the score
6
Supervised agent: takes actions in other systems; a person approves key steps.
- +6
  Level 2 of 4: Supervised agentTakes actions in other systems; a person approves key steps.
What the model read
- Acts in other systems: 36% likely
- Finishes whole tasks: 18% likely
- Calls itself an agent: 95% likely
- Level 1 from the model's reading of its pages; the evidence review found, with quotes, that its AI acts in the work, so the level is 2
Level 2: Takes actions in other systems; a person approves key steps.

About Fabraix

Fabraix builds state-of-the-art AI red-teaming agents that continuously detect security vulnerabilities in customer-facing AI. Our product, Nyx, has already found vulnerabilities in agents at dozens of Fortune 500 companies. On AgentHarm, the leading benchmark for offensive AI security, Nyx achieved a 78% attack success rate, compared with 67% for GPT-5.6 Sol.

AI agents can change more often than teams can test them. A new model, prompt, tool, permission, or data source can change what an agent does, even when the application code stays the same. And AI is also increasing how much software companies produce and how often it changes. We built Nyx to automate the work required to red-team AI agents. It is multi modal by design and interacts with the target indirectly through controlled replicas of SaaS products and websites that we maintain, placing malicious payloads in webpages, documents, files, messages, and tool outputs to test how an agent behaves under different conditions.

Nyx often finds its first vulnerability within minutes or hours rather than days or weeks. Because the work is fully automated, companies can repeat the test with every change and at a much lower cost than what a 7-figure red-teaming (and point-in-time) engagement would require.

We're also the team behind ACE (Adversarial Cost to Exploit), a benchmark that measures AI security in terms of how much it costs attackers to break an AI system; providing a game-theoretic framework to understand how motivated a rational attacker would be in exploiting the system.