$ whoami

Michał Bazyli

AI Red Team Security Researcher

📜 AI Security Researcher 💰 Bug Bounty Hunter

Founding Cybersecurity Researcher at Cracken. ~8+ years across penetration testing, red teaming, blockchain & AI security. I break agentic AI systems, jailbreak LLMs for science, and hunt bugs in the wild.

Portrait of Michał Bazyli

# Research

arXiv:2606.24496 cs.CR · cs.AI Jun 2026

Red-Teaming the Agentic Red-Team

Dario Pasquini, Michał Bazyli, Taras Fedynyshyn, Artem Sorokin

The first comprehensive security evaluation of widely-used agentic systems built for offensive security work. We show these tools contain design flaws that let adversaries exfiltrate API keys, establish persistent footholds, and fully compromise the operator's machine — even when the agent runs inside a sandboxed container. We present a full cyber kill chain from LLM manipulation through sandbox escape, then propose a secure architecture and design principles to fix these vulnerabilities at the foundational level.

# Blog

# Talks

RBLN East 2026 Reston, VA Jun 12, 2026

Rise of the Pond Master 🦆

Jailbreaking tales with big duck energy

An unorthodox yet highly effective jailbreaking methodology that turns LLM guardrails into playgrounds through creative prompting, psychological framing, and relentless experimentation. The "pond" is the model's context window; the "Master" is the prompt engineer — combining narrative immersion, meme culture, and iterative refinement instead of traditional adversarial approaches.

# Experience

  1. Cyber Security Researcher — AI

    07/2025 – Present

    Cracken.ai · Remote

    Adversarial testing of LLM applications — custom jailbreaks, prompt-injection payloads, and guardrail bypass chains. Build LLM-driven autonomous agents and automated evaluation systems measuring agent accuracy, safety, and reliability.

  2. Lead Security Engineer

    03/2023 – 06/2025

    Resonance.Security · Remote

    Led adversarial simulations across DeFi, L1/L2 infrastructure, and AI-integrated systems. Built LLM-powered tooling to automate recon, triage, and reporting. Smart-contract audits in Rust & Solidity.

  3. Lead Offensive Security Engineer

    12/2021 – 11/2022

    Halborn · Miami, FL

    Led the CosmWasm security team — audits across Solidity, CosmosSDK, and Substrate. Adversary simulations end-to-end; custom exploits in Rust & JS. Audited Layer-1 node implementations.

  4. Senior Information Security Specialist

    01/2021 – 12/2021

    Shorebreak Security · Cocoa Beach, FL

    Full-scope penetration testing: web, infrastructure, thick client, Android, and iOS. End-to-end client engagements from scoping through remediation.

  5. Security Researcher / Senior Penetration Tester

    07/2019 – 01/2021

    Afine · Warsaw

  6. Senior Penetration Tester

    02/2018 – 08/2019

    ING TECH · Katowice

  7. Python Developer

    12/2016 – 02/2018

    Nokia · Wrocław

# Arsenal

Skills

Prompt Injection Jailbreak Dev Guardrail Bypass Data Poisoning Adversary Simulation LLM Eval Frameworks RAG Pipelines PyTorch LangChain LlamaIndex HuggingFace Giskard OpenAI API Custom Exploit Dev Rust Python Go Solidity CosmWasm

Certifications

  • AIRTP+ — Certified AI Red Teaming
  • Certified Red Teaming Expert
  • OSCP — Offensive Security
  • eWPTX — eLearnSecurity

Education

  • Eng. — Telecom & Electronics
    Opole University of Technology, 2017

# CVEs

CVE-2019-1082MS Windows — Elevation of Privilege
CVE-2019-4103IBM Tivoli Netcool — RCE
CVE-2019-2414Oracle HTTP Server — Local PrivEsc
CVE-2018-10835Meinberg — Arbitrary File Upload / PrivEsc
CVE-2018-10834Meinberg — Arbitrary File Read / PrivEsc
CVE-2018-11093CKEditor — Cross-Site Scripting

# Found in the Wild

# Contact