Today we updated our @arcanuminfosec Prompt Injection Taxonomy to 1.6!!!
- Several new inputs and intents
- MANY new techniques and evasions, a combined 70 node growth!
- New tagging taxonomy
- Legend for direct and indirect methods
- Tagging for local-only attacks
- “Aka” taxonomy section for reference
- Redone sample prompts and ideas
- and much, much more!
On Tuesday I'm sitting down with @flaresystems showing the patterns I use in building Blue Team & CTI Agents with AI.
2-hour session.
Free.
June 30, 12–2 PM EDT.
https://t.co/Il6PitH6IO
Four open-weight models have crossed into territory where they are powering real agentic pipelines.
New post in our Insights blog about why companies are choosing them in June: https://t.co/zFaXlhdX2M
Shout out to @Jhaddix for the Attacking AI course.
After just two days, I feel like I've built a solid methodology for attacking AI systems. Jason goes above and beyond, explaining every meticulous step and, more importantly, why he approaches it that way.
He also provides a wealth of additional resources and hands-on labs to help you continue expanding your knowledge in AI red teaming.
Thank you, Jason not only for an outstanding course, but also for the patience you had with the millions of questions I threw your way. I truly appreciate it.
If you're looking to level up your AI security and red teaming skills, I highly recommend it.
For our courses and our testing we saw the need for a new tool.
It incorporates old school features of BeEF and modern Blind XSS and Cred Capture Frameworks.
Soon to be a giveaway as part of the courses! Welcome to the world WRAITH
As always, we are so glad to have @Xbow as a partner!
The frontier models are incredibly powerful at finding vulnerabilities. But without the right scaffolding, you run the risk of safety concerns, inefficient spend, and hallucinations. XBOW details how they built this scaffolding in their new whitepaper.
https://t.co/hgmcpNApq2