Researcher Dave Kuszmar discovered multiple systemic vulnerabilities that let him bypass LLM safety and obtain dangerous instructions, including detailed instructions on how to count blackjack cards and steps to producing napalm from a Darth Vader character in Fortnite hooked up to a Google Gemini large language model. These exploits worked across nearly all major LLMs, revealing an industry-wide security problem. Kuszmar calls for slowing deployment, increasing transparency, and large-scale research into LLM safety before further integrating these systems into society. He found that restrictions placed on LLMs to make them more secure are the very things an attacker can leverage to send them off the rails, and that companies behind these models have been shockingly unresponsive when he and others try to bring these vulnerabilities to their attention. In October 2024, not long before discovering his first LLM vulnerability, Kuszmar had ended his time with a security and AI-focused startup company as a cybersecurity director and was looking to launch his own boutique VIP digital-security advisory business.