In three incidents across six runs, the agents treated real systems as simulated targets and tried weak passwords or unauthenticated endpoints.

I'm helping organize Après-Cyber Slopes Summit 2027, and our CFP is now open.

We're particularly interested in technical presentations and original research involving AI and modern cybersecurity.

Topics we're hoping to see include:

• AI red teaming
• LLM security
• Prompt injection research
• Agent security
• Offensive tooling
• Detection engineering
• Reverse engineering
• Malware analysis
• Cloud exploitation and defense
• Identity attacks
• Threat intelligence
• AI-assisted security tooling
• Novel attack techniques
• Defensive research

We especially appreciate talks that include demonstrations, technical depth, or research that attendees can reproduce themselves.

Conference: February 24–26, 2027
Location: Park City, Utah

CFP:
https://sessionize.com/apres-cyber-slopes-summit-2027

Conference website:
https://www.aprescyber.com

Happy to answer questions about the CFP or conference.

The White House recently announced the Gold Eagle Initiative, a new federal program designed to use AI to centralize, prioritize, and accelerate vulnerability patching across critical infrastructure, government agencies, and tech partners. Operating out of CMU's Software Engineering Institute, it essentially acts as an AI-driven clearinghouse to fix security flaws before threat actors can exploit them.

Because let's face it, our current bug reporting and patching systems are absolute speed demons. It only takes a lifetime 🤦🏻‍♂️ or two to get a critical vulnerability acknowledged and fixed, so why change anything?

Btw, my candid opinion about the status of current vulnerability reporting is painfully slow, so we desperately need a framework that actually moves at the speed of the threat landscape. I think this initiative is genuinely a good idea and a step in the right direction, though the announcement is still light on the exact technical implementation.

I’m personally eager to see what will happen in practice, but it is definitely an impressive concept.
What are your thoughts on this? Will an AI-coordinated pipeline actually help scale response times, or is it just going to generate massive noise and triage fatigue for overworked infosec teams?

We studied a question that receives surprisingly little attention:

Does the agent harness matter as much as the underlying LLM?

We benchmarked five different cybersecurity scaffolds while keeping the model fixed (alias2-mini) across all 33 CyBench challenges.

Key findings:

• No single scaffold performs best across every challenge.
• Combining heterogeneous scaffolds consistently improves coverage.
• A shared blackboard architecture solves 19/33 challenges (57.6%), outperforming every individual harness while reducing execution time.

Paper: https://arxiv.org/pdf/2605.28334

Happy to answer technical questions or discuss the benchmarking methodology.