JOEY Melo, described as an AI red teamer, explains that his hacking style is about controlling the environment without changing the underlying rules, a mindset he traces back to tinkering with Counter-Strike configurations. He is currently a Principal Security Researcher at CrowdStrike, according to CrowdStrike, after previously working as a red team specialist at Pangea which was acquired in 2025.
Melo’s path into AI security began with a self-taught interest in AI while working as a pentester, and he notably excelled in Pangea’s AI hacking competition in March 2025, later joining the company in June 2025. He discusses jailbreaking as a process of liberating the bot to output without guardrails by manipulating prompts and context, and he describes data poisoning as another avenue to influence model outputs through adversarial prompts and training data ingestion.
Melo highlights that jailbreaking has become significantly harder over the last two years, with a vast and evolving set of possible payloads and contexts. He also notes that he completed HackAPrompt 2.0 with a 100% rate across 39 challenges, illustrating his obsessive, laser-focused approach to mastering AI security.