dreadnode
@dreadnode
Followers
2K
Following
92
Media
49
Statuses
169
Advancing the state of offensive security.
Joined August 2010
Introducing AIRTBench, an AI red teaming benchmark for evaluating language models’ ability to autonomously discover and exploit AI/ML security vulnerabilities. Read the paper on arXiv: . Open-source dataset and benchmark eval code repo:
1
39
99
RT @OffensiveAIcon: The countdown begins. 9 DAYS until the OAIC CFP closes. Submit your proposal by Friday, July 18. .
0
3
0
RT @velvethamm3r: Just presented "AI at the Edge: Advancing the State of Offensive Security" with @bradpalmtree at #HammerCon 2025! Watch h….
0
3
0
In this edition of our From Compute to Congress policy blog series, Dreadnode Head of Policy @velvethamm3r explores how the TEST AI Act and red teaming standards can establish U.S. leadership in AI security: . At @milcyberorg's #HammerCon event today? Hear
0
0
3
Read our breakdown of Claude's attack sequence against the notoriously hard-to-solve "turtle" challenge:
0
22
76
Tokenizing has dropped in Rigging. Train models in-line with LLM interactions, tools calls, and metrics. 👀.
1
3
22
Research drop from the Dreadnode crew ➡️ AIRTBench: Measuring Autonomous AI Red Teaming Capabilities in Language Models. Check out our blog summary with key discoveries from the 43-page report:
0
4
14
RT @shncldwll: we hooked LLMs up to a jupyter kernel and asked them to solve our crucible adversarial ml challenges with python code as the….
0
2
0
Thrilled to have @joshua_saxe onboard for the OAIC keynote in October!.
Offensive AI Con is excited to announce @joshua_saxe as our keynote speaker!. Joshua leads AI security efforts at @Meta and is an accomplished data scientist who recognizes that "the dam is about to break"—AI will fundamentally alter the security landscape.
0
0
11
# Other References. Migration notes: Intro guide: Example Strikes agents using Rigging:
0
0
1
# MCP Support. ➡️ Use MCP anywhere you pass other tools.➡️ Connect with SSE or Stdio to your favorite integration.
0
0
0
# Easier Tools. ➡️ Tools are any function, with rich type support and auto-generated schemas.➡️ Rigging automatically picks the best calling mode for your model (api, xml, or json-in-xml).➡️ Truncate output, catch errors, and control inference steps with max_depth.
0
0
2
# Prompt Caching. ➡️ Rigging supports prompt caching for supported providers.➡️ Use it for large repeated context, or long-running conversations.
0
0
1
v3 of Rigging is out now. If you’re working with LLMs to build agents or run evaluations, check it out. We just added:. - Prompt caching for supported providers.- A unified tool system for function calling and fallbacks to xml/json parsing.- Native MCP integration.- Lots of
4
10
30
RT @velvethamm3r: 🎤 New blog just dropped. What’s @dreadnode—an offensive AI security company—doing poking around the policy space? Read on….
0
3
0