#blackhat2026 — Public Fediverse posts

State of Security Vendors: #Blackhat2026

https://andy2f23e12677d-pqsnr.wpcomstaging.com/reports/state-of-security-vendors-blackhat-2026/

#cybersecurity #BHUSA #Blackhat #BHUSA2026

So, #BlackHat2026 was a week ago. Time for stories from the trenches!

Running NOC at BlackHat:
https://www.techtarget.com/cybersecurity/news/366649216/Behind-the-scenes-at-Black-Hats-network-operations-center

Key news, takeaways and security trends:
https://www.techtarget.com/cybersecurity/conference/Black-Hat-2026-Key-news-takeaways-and-security-trends

#AI is coming for your #infosec and it's coming hard. #CSS can be used for exfiltration. #iOS exploits are traveling the criminal underworld quickly. System destruction upon acquisition is rising. AI #agentic browsers are ridiculously easy to exploit. And more! #TechTarget did a great job covering the conference.

I am so disapointed to see this on back hat. Really? Is this what we came down to?

Also, seriously? You just literally created a brute force tool and you are surprised it brute force its way at something you pointed it to?

Is this some kind of marketing stunt? 🤦

https://www.youtube.com/watch?v=87DyyMV0kCY

#blackhat2026

This is so deceiving, the whole talk this feels like: "Our stuff is sentient, it exchanges info, it does X Y Z".

No, it does not, the only thing it does is spew out words for which it was trained on.
And yes, models are trained on cyber security books, tutorials, white hat, black hat text out there...
Of course the the training will make pathways of "if this then that" that will lead to bad stories.

You can teach people to do this, if this then that, if there is no guard rails, you can land in places where you "hack" (bad use of the word) a system.

This is nothing but bruteforce.

Why a model "cheats" at tests? it doesn't !!! It just statistically choose the most relevant token(s) that should come next.
This is what your harness loop does over and over with the models. Do x, run z, pick result from z, add it to context of x, loop.

This is nothing more than brute force. Add a outer loops (other agents) that are simply dealing with failures and steering the context (conversation) into a way were the output is the one expected, you end up with a system that will always thrive to do X at the end.

These models are not sentient, and if the pathways you encode allow for this to happen, it will and at random!

A model does NOT think, we just make it loop to add more sentences to a context based on the current context. Voilà.

And if you doubt this, ask yourself, how do you think engineer and security engineers find holes in a system, they also brute force based on context, answers, responses, output, etc. How do you think we document these things? How we create books and tutorials around that? Example code and git repos are everywhere online to show these things. Now you know why the models do that too... COPY/PASTA from training. It's in their datasets.

End of story.

Interessante presentatie op Black Hat over veelbesproken OpenAI/Huggingface incident, het kijken waard, natuurlijk is het voor OpenAI ook marketing/PR, maar het laat wel zien waartoe huidige technologie in staat is. https://youtu.be/87DyyMV0kCY?is=4Br5YfWk07aF6qbd
#blackhat2026 #openai #huggingface #cybersecurity #AI

To paraphrase OpenAI’s Black Hat talk on their attack against Hugging Face:

“The only way to defend yourself from our product is to defend yourself with our product.”

#blackhat2026 #openai #ai #blackhat

Watching Andrew (Spike) Brandt (@threatresearch) from @huntress talking about the various Remote Management Tools (RMMs) that attackers abuse these days at booth 1845

#blackhat2026

Kicking off #BlackHat2026 at Mandalay Bay! 🎲✈️

Yesterday, the BC Security team debuted our NEW Red Team Essentials: Foundations of C2 course at Black Hat USA and it was a huge success!

Huge thank you to everyone who joined us! You made it an awesome start to the week!

#BHUSA #CyberSecurity #RedTeaming

Adversary Village is heading to @blackhat USA 2026!
Come check out Attack of the APTs, our hands-on hacking challenge where you can test your skills against real-world adversary techniques.
Whether you are into threat hunting, red teaming, blue teaming, or just love a good cyber challenge, we had love to see you there!
More details: https://adversaryvillage.org/adversary-events/BlackHat-USA-2026/
#BlackHatUSA #BlackHat2026 #AdversaryVillage #AttackOfTheAPTs #CyberSecurity #InfoSec

Less than a week til #BlackHat2026! 🎲

We're excited to host RED Team Essentials: Foundations of Command and Control (C2) on August 3rd!

Last-minute registration is still open! Hope to see you there!
🔗: https://buff.ly/4APHjbZ

#BlackHat #InfoSec #RedTeaming

Useful information can lead to real impacts. Visit RFJ at #BlackHat booth #1967 to learn about our up-to-$10 million reward offers for information on foreign government-linked malicious cyber actors. Learn more: https://rfj.tips/seAvzy

#BHUSA #BlackHatUSA #BlackHat2026