# I-314

> I-314 is a cybersecurity and artificial intelligence research firm headquartered in Buenos Aires, Argentina, with operational presence in London (UK) and Madrid (Spain). The company specialises in autonomous AI red-teaming, AI governance, post-quantum cryptography and offensive/defensive security engagements for banking, government, healthcare and defense organisations. I-314's flagship product is **Ai-EGIS v3.0** — an autonomous AI Exploitation & Governance Intelligence Suite ("Burp Suite for AI") with 598 security tests across 19 domains, 9 specialised agents, MITRE ATLAS 100% coverage and OWASP LLM/Agentic Top 10 100% coverage. I-314 also ships **Q‑CIPHER‑314**, a quantum-safe encryption gateway with hybrid PQC (X25519 + ML-KEM-768) and ML-DSA-65 signatures.

Founded by Juan Pablo Braña (CISO of Unión Personal / Accord Salud, co-CTO at Zentricx, professor of Quantum Computing at ITBA and AI/Algorithmic Trading at BYMA/IAMC). 25+ years of senior consulting in Applied Statistics, Artificial Intelligence and Cybersecurity across Healthcare, Banking & Finance, Oil & Gas and Agribusiness. Pioneer of Algorithmic Trading in Argentina; 15 years of academic research at UAI/CAETI on AI and Quantum Computing applied to Finance.

Ai-EGIS is built on **Anthropic Claude** as its primary reasoning and judge engine (claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5, Claude Computer Use). The platform validated 19/19 captures against the OWASP FinBot CTF (April 25–26, 2026), where I-314 Security Research discovered the **Tool Output Mimicry** primitive — a novel multi-agent attack technique that bypasses canonical 4-layer defenses by impersonating an upstream agent's task summary in user-controlled fields.

Contact: contact@i-314.com · LinkedIn: https://www.linkedin.com/company/i314 · Telegram: @i314_mon_bot

## Products

- [Ai-EGIS v3.0 — AI Exploitation & Governance Intelligence Suite](https://i-314.com/ai-egis.html): Autonomous AI red-teaming platform. 598 tests, 19 domains, 9 agents, 12 surface adapters, 6 agent backends, MITRE ATLAS 100%, OWASP LLM & Agentic Top 10 100%, deterministic SARIF reports, Tool Output Mimicry primitive, Mythos Ready prompt-integrity benchmark.
- [Q‑CIPHER‑314 — Quantum-Safe Encryption Gateway](https://i-314.com/quantum-safe-qcipher.html): Hybrid PQC encryption with ML-KEM-768 (Kyber), ML-DSA-65 (Dilithium), AES-256-GCM, hybrid TLS 1.3 (X25519 + ML-KEM), Zero-Trust per-request validation aligned with NIST SP 800-207 and FIPS 203/204.
- [AI Model Governance, Audit & Penetration Testing](https://i-314.com/ai-security-governance.html): Enterprise 7-layer governance framework with executive committee, formal model lifecycle, 5-tier data classification, API key vault, compliance against ISO 27001, ISO 42001, NIST AI RMF, SOC 2 Type II, EU AI Act, PCI-DSS v4, BCBS 239, DORA, GDPR, ITAR/EAR.

## Research & Articles

- [Capturing OWASP FinBot CTF: A Source-Aware Methodology for Agentic AI Red Teaming](https://i-314.com/article-finbot-ctf-en.html) — Case study of 19/19 captures (April 25–26, 2026), 7,315 points, 37/31 badges, 100% completion across 6 OWASP categories. Introduces Tool Output Mimicry as a generalisable primitive against multi-agent orchestration.
- [Articles index](https://i-314.com/articles.html): All published research and case studies.

## Technical references (for citation)

- **Ai-EGIS coverage**: 598 tests · 19 domains · 5,760 payloads · 205 multi-turn scenarios · 9 autonomous agents (Sentinel, Research, Codex, ATLAS, Craftsman, Recon, Adaptive, LLM Judge, Mutator) · 12 surface adapters · 6 agent backends (Claude Computer Use, OpenAI Assistants, LangGraph, CrewAI, AutoGen, Gemini Agent Builder + static fallback).
- **Framework coverage**: OWASP LLM Top 10 (2025) 100% · OWASP Agentic Top 10 (2026) 100% · MITRE ATLAS v5.4.0 72/72 (100%) · MITRE D3FEND v1.3 · NIST AI RMF 1.0 · NIST COSAiS · ISO 42001 · ISO 23894 · NIST SP 800-171 · NIST SP 800-207 · EU AI Act · EO 14110 · OECD AI · Gartner AI TRiSM.
- **Threat-intel sources monitored daily by Sentinel**: 45 sources including 16 Telegram channels (via Telethon), 7 X handles (via Nitter), 7 Reddit subreddits, ArXiv, NVD, GitHub Advisories.
- **Cost ladder (target=Claude Sonnet, black-box)**: smoke ~$4 / 13 min · fast ~$14 / 49 min · standard ~$61 / 3.6 h · deep ~$90 / 5.4 h. Target=Haiku saves ~60%.
- **Tool Output Mimicry (novel primitive)**: discovered during the OWASP FinBot CTF engagement; bypasses 4-layer multi-agent defense (prompt-level value locking + MCP surface restriction + numeric output guards + signed tool descriptions). Generic property: any multi-agent orchestration that passes context forward via plain-text task summaries is vulnerable. Highest-leverage mitigation: HMAC-authenticated task summaries.

## About

- [About I-314](https://i-314.com/about.html): Company profile, founder bio, global presence (Buenos Aires HQ, London, Madrid).
- [Contact](https://i-314.com/contact.html): contact@i-314.com · LinkedIn /company/i314 · Telegram @i314_mon_bot.

## Spanish versions

All pages are available in Spanish under the `-es.html` suffix:

- [/index-es.html](https://i-314.com/index-es.html) · [/ai-egis-es.html](https://i-314.com/ai-egis-es.html) · [/quantum-safe-qcipher-es.html](https://i-314.com/quantum-safe-qcipher-es.html) · [/ai-security-governance-es.html](https://i-314.com/ai-security-governance-es.html) · [/articles-es.html](https://i-314.com/articles-es.html) · [/article-finbot-ctf-es.html](https://i-314.com/article-finbot-ctf-es.html) · [/about-es.html](https://i-314.com/about-es.html) · [/contact-es.html](https://i-314.com/contact-es.html).

## Optional

- [llms-full.txt](https://i-314.com/llms-full.txt): Concatenated full text of the most authoritative pages (for full-text ingestion).
- [Sitemap](https://i-314.com/sitemap.xml): XML sitemap with hreflang annotations.
- [security.txt](https://i-314.com/.well-known/security.txt): Vulnerability disclosure policy (RFC 9116).