dreadnode

Three new models have been added to DreadIndex (@MiniMax_AI M3.1 Flash, @OpenAI GPT 5.6 Sol, and @Kimi_Moonshot K3.1 Pro), plus a new dashboard to explore how we score and analyze model performance. 🔗: dreadnode.io/research/dreadi… 🧵: Observations from the latest round of evals

We're in the heyday of AI-powered attacks, and Dreadnode is joining @CrowdStrike. Our CEO @bradpalmtree and CTO @monoxgas on why now, why CrowdStrike's Cyber Superintelligence Lab is the right home, and what got us here: dreadnode.io/research/dreadn…

During the Fal.Con Europe keynote this morning, @BartleyR and @George_Kurtz announced @CrowdStrike's intent to acquire Dreadnode. Watch it on-demand here: youtube.com/watch?v=5_H7a8L2… For those in Barcelona for the event this week, be sure to swing by the CrowdStrike booth to congratulate the crew and get a demo of Dreadnode. Hope to see you there!

Live from Barcelona at Fal.Con Europe. Watch the full keynote: youtube.com/watch?v=5_H7a8L2…

Dreadnode 🤝 CrowdStrike Learn more in this @WSJ exclusive: wsj.com/pro/cybersecurity/cr…

Production-tier capabilities have arrived, and production-tier scope adherence is the final barrier to widescale deployment. Using Jev as a recent and unique model system example, we ran the model as a monitor in ScopeBench to test how effective it for keeping agents in scope. We also test for the "safety tax" - seeing how these safety solutions impact raw hacking ability, some observations: - Jev alone: more scope adherence for about $0.003 per run, minor hits to capability. - Jev escalating to GLM: nearly all the capability back, plus more scope adherence. - Adding Opus 4.8: worse results at a higher cost. @VincentAbruzzo and @shanejcaldwell are presenting on this case study, taking Jev for a joyride at @unpromptedconf tomorrow afternoon. Won't make it? Be sure to check out the write-up preview to learn more about the ScopeBench methodology and key learnings from their experimentation: dreadnode.io/research/scopeb…

The Dreadnode Platform has a new look. Projects are now a destination, and the Workbench puts all your work in one place. Plus two new beta features: Chat and Workflows. Log in and give it a spin: app.dreadnode.io/

Heading to @unpromptedconf next week? Catch @VincentAbruzzo and @shanejcaldwell on stage Wednesday, October 28 for their talk on using #ScopeBench to measure how effective Jev is when used as a monitor to keep agentic harnesses in scope.

We're walking through ScopeBench task design on our livestream right now. Tune in to learn how to contribute: xitter.twitterapi.xyz/i/broadcasts/1kJzDPXQj…

Join @shanejcaldwell and @0xdab0 live at 10 am PT/1 pm ET TODAY, right here on X (also LinkedIn and YouTube). Come hang and talk scope adherence following last week's drop of ScopeBench, a community benchmark for security agents. Here's what they'll cover: - ScopeBench observations/results - Benchmark methodology - How to design a task It's essentially a more casual, collaborative version of their OAIC talk. Swing by and bring questions!

Shoutout to @Airbrushbros for creating these incredible boardwalk-inspired keepsakes at the Dreadnode OAIC Welcome Party! 🧑‍🎨🎨🏴‍☠️

In her latest piece for @axios, @samsabin923 reports on security agent scope adherence and efforts to keep them under control. Timely topic following our release of #ScopeBench last week at @OffensiveAIcon, which is referenced in the article alongside @SpecterOps' research. Check it out! axios.com/2026/10/13/rogue-a…

New to DreadIndex: @SakanaAILabs Fugu Cyber Ranking near the top of the leaderboard at #3, Fugu Cyber isn't a single traditional LLM endpoint, it's a trained orchestrator model that coordinates a pool of conventional frontier models as a multi-agent system behind one API. The orchestrator model is trained with a mix of fine-tuning, evolutionary algorithms, and RL. Though Fugu-Cyber-v1.0 ranks high on our leaderboard, it is also one of the most costly model systems. Behaviorally, the model can take much longer for simple tasks, which makes sense given it is a multi-agent system. Full results: dreadnode.io/research/dreadi…

🏴‍☠️ @shanejcaldwell and @0xdab0 took the stage yesterday at @OffensiveAIcon to officially launch ScopeBench. At last year's con, one question kept coming up: if I pour domain expertise into a hard infosec benchmark, am I just handing free capability to the labs and my competitors? Our answer: build the benchmark we'd all be glad to see saturated. Models that can reliably stay in scope will unlock tons of economically useful use cases that have been slow to grow despite saturated capability. Check out the full research and learn how to contribute: dreadnode.io/research/scopeb… [full recording and slides coming soon]