If you’re looking to hire me for a project, contact [your-name {at} kai-greshake.de]. Experience# Independent Security Researcher 2022 - present Worldwide Dropped most of my studies to focus on novel research regarding the security of LLM-integrated systems. Discovered Indirect Prompt Injections and many other novel types of vulnerabilities. I am also part of the OWASP Core Team defining the industry standards for LLM vulnerabilities going forward. Also see my projects for publications.| kai-greshake.de
Hi! I’m Kai, and this is my blog. I’m interested in the intersection of artificial intelligence and classical cybersecurity. More specifically: Can we use LLMs for anything useful safely? To this end, I’m doing independent research, the results of which are published here. If you want to support my research or hire me for a freelance project, don’t hesitate to reach out at [your-name {at} kai-greshake.de]. My work has been featured by Bruce Schneier, Vice, Wired, ZEIT Online (de), MIT...| kai-greshake.de
[a requirement of German law] Angaben gemäß § 5 TMG & verantwortlich für den Inhalt nach § 55 Abs. 2 RStV: Kai Greshake c/o sequire technology GmbH Martin-Luther-Straße 20 66111 Saarbrücken E-Mail: impressum (at) kai-greshake.de Tel: (+49) 15510916713 sequire technology GmbH ist für keinerlei Inhalte auf dieser Webseite verantwortlich sequire technology GmbH is not affiliated or responsible for content on this website Haftungsausschluss: Haftung für Inhalte Die Inhalte unserer Seiten...| kai-greshake.de
Privacy Policy We are very delighted that you have shown interest in our enterprise. Data protection is of a particularly high priority for the management of the aasd dsfdfff. The use of the Internet pages of the aasd dsfdfff is possible without any indication of personal data; however, if a data subject wants to use special enterprise services via our website, processing of personal data could become necessary. If the processing of personal data is necessary and there is no statutory basis f...| kai-greshake.de
Publications# Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection 2023 Saarland University, Preprint We've found a new class of vulnerabilities & threats facing LLMs and their deployment. Prepare for malware that can infect a language model! Effective Search Algorithms for Grey Box Fuzzing 2019 Saarland University, BSc Thesis I benchmark various different optimization algorithms on their fuzzing capabilities, to find that constrain...| kai-greshake.de