New service · AI Security

Ship AI. Not Vulnerabilities.

Ship AI. Not Vulnerabilities.

Ship AI. Not Vulnerabilities.

Independent security audits for LLM apps, AI agents, RAG pipelines and open-source models — built for finance, research and data teams.

Independent security audits for LLM apps, AI agents, RAG pipelines and open-source models — built for finance, research and data teams.

Book a free scoping call

Book a free scoping call

See what we audit

See what we audit

Your AI is already a new attack surface

Your AI is already a new attack surface

Your AI is already a new attack surface

Prompt injection, data leakage through RAG, over-privileged agents and unverified open-source models don’t show up in classic pentests. We find them before attackers, auditors or clients do.

✅ Map every AI component, data flow and integration

✅ Hands-on testing of models, agents and tools

✅ Prioritized fixes your engineers can ship this sprint

✅ Evidence for compliance, clients and the board

Scope

What we audit

What we audit

AI Architecture Review

Data flows, trust boundaries, model hosting, secrets and access control across your AI stack.

LLM & Agent Red Teaming

Direct and indirect prompt injection, jailbreaks, tool abuse and excessive agency.

RAG & Data Leakage

Permission-aware retrieval, vector DB exposure, PII leakage and data poisoning.

Integrations, APIs & MCP

Connectors, plugins, MCP servers, OAuth scopes and third-party AI APIs.

Open-Source & Model Supply Chain

Model provenance, unsafe weights and serialization, dependency CVEs and licenses.

Governance & Compliance Readiness

Mapped to OWASP Top 10 for LLM & Agentic apps, NIST AI RMF, ISO/IEC 42001 and the EU AI Act.

Process

How it works

How it works

01

Scope

Free 30-min call. We map your AI use cases and agree on scope.

02

Assess

Threat model plus hands-on testing in your environment.

03

Report

Executive summary and prioritized findings with proof of concept.

04

Fix & Retest

We help your team remediate and verify every fix.

Packages

Choose your engagement

Choose your engagement

AI Risk Snapshot

1 week

  • One AI app or use case

  • Architecture & threat review

  • Top risks + executive summary

  • Best before launch

Book a call

Most popular

Full AI Security Audit

2–4 weeks

  • Architecture + red teaming

  • RAG, integrations & supply chain

  • Compliance mapping

  • Retest included

Book a call

Continuous AI Assurance

Monthly

  • Testing of new models, prompts & agents

  • Security gate before each release

  • Direct access to our experts

  • Quarterly risk report

Book a call

Why AQURIX

Why AQURIX

We build AI ourselves

We ship local LLMs, RAG and agents, so we know exactly where they break.

We build AI ourselves

We ship local LLMs, RAG and agents, so we know exactly where they break.

Finance-grade confidentiality

NDA by default. Testing runs in your environment; your data never leaves it.

Finance-grade confidentiality

NDA by default. Testing runs in your environment; your data never leaves it.

Actionable, not a 200-page PDF

Clear risk scores and fixes your team can implement immediately.

Actionable, not a 200-page PDF

Clear risk scores and fixes your team can implement immediately.

FAQ

FAQ

Do you need access to our production data?

Do you test apps built on OpenAI, Anthropic or Google models?

How long does an audit take?

Will this help with the EU AI Act?

What happens after the audit?

Know your AI risk before attackers do.

Know your AI risk before attackers do.

Book a free 30-minute scoping call. NDA from the first conversation.

Book a free scoping call