Amidst the craziness in AI and cybersecurity this month, things continue to progress. Most notably, OWASP released the second version of their LLM Top 10 list and it's more useful for businesses adopting AI because the initial version was very focused on the issues with LLM models themselves, which is the responsibility of the OpenAI's of the world.
This new list looks at how LLM features and applications are built in the real world so it prioritizes things we've been talking about, like the weaknesses with vector embeddings and RAG workflows.
Check out our blog on the topic or go straight to the source yourself. It feels good to have an organization like OWASP validating what IronCore's been saying for the last 18 months.
Stay safe and protect your data.
Patrick Walsh CEO, IronCore
OWASP's Updated Top 10 LLM Includes Vector and Embedding Weaknesses
The Update Looks Beyond Models to the Whole AI Stack
OWASP released their second version of the Top 10 for LLM Applications. It now includes major new issues found in the surrounding AI ecosystem, going beyond LLM model risks. In this blog post, we look at their key findings and zoom in on LLM08, vector and embedding weaknesses.
This twelve minute video covers embeddings, vector inversion attacks, how they work, what can go wrong, and how to use encryption to stop them. This is intended as a shorter briefing video for those just learning about the subject.
