Application-oriented language models exhibit threats due to intrinsic vulnerabilities. In February 2023, we discovered this in collaboration with CISPA – Helmholtz Center for Information Security. Now, the Federal Office for Information Security (BSI) has also published a cybersecurity warning regarding the security gap we found, cautioning against the risk posed by Indirect Prompt Injections when integrating LLMs (Large Language Models) into applications.

To access the publication, please follow this link: https://www.bsi.bund.de/SharedDocs/Cybersicherheitswarnungen/DE/2023/2023-249034-1032.pdf?__blob=publicationFile&v=3.