@SecurityBreakAI

Threat Intelligence for the AI Ecosystem

Joined July 2026
  • Tweets24
  • Following53
  • Followers208
13Media
Your AI agents are under attack. Our founder @fr0gger_ is in San Francisco for 2 weeks for @startmate Winter ‘26 Demo Day to talk about it. Investors, security leaders, AI security folks: want a closer look at what we’re building? DM us and let’s meet.
1
1
25
Your org is running more AI than it can see and attackers are exploiting it. Watch @fr0gger_ talk on AI Threat Intel & build your defenses at our upcoming trainings: • Black Hat Europe (Dec 7-8) • Black Hat Asia (Feb 28-Mar 1) Links in comments👇 youtu.be/hqCO0S8qzVA

[un]prompted.au 2026 -AI Threat Intelligence: Hunting Threats Across the AI Ecosystem -Thomas Roccia

AI is transforming how we work, how we build software, and how we i...

youtube.com
1
3
3
170
🚀 Black Hat Asia 2027: blackhat.com/asia/training/s… 🔥 Black Hat Europe 2026: blackhat.com/europe/training…
1
3
51
Big night ahead, Demo Day is here! 🎉 We’re presenting NovaHunting tonight as part of the Winter 26 @startmate cohort. Massive thanks to everyone who helped us get ready over the last few months! 📍 The Timber Yard, Melbourne 🗓️ Tonight @ 6PM 🎟️ Tickets: luma.com/DemoDay-W26
1
1
81
Melbourne, @fr0gger_ is in town! 🇦🇺 After his #CyberCon talk today, he's staying until @startmate Demo Day on 19 Oct 🚀 Security teams, founders, investors: want to meet him and see NovaHunting in action? DM us for a coffee or a demo ☕ #AIsecurity #NovaHunting #ThreatIntel
1
1
1
2
1,100
AI agents read data, call tools, and take action, making them prime targets for attackers. NovaHunting stops prompt injections, hijacking, and data leaks while keeping you in total control: ⚙️ Custom detection rules 👁️ Full visibility into actions 🔌 Integrates with your stack
1
2
88
Meet us at #CyberCon2026 Melbourne. @fr0gger_ will be sharing practical insights on AI threat intelligence. 🎤 “Your AI threat intelligence is broken. Here is what works.” 📅 Thu 15 Oct | 4:40–5:20 PM 📍 Melbourne
1
1
3
1,110
Your AI agent has access to your data... but who else is giving them instructions? 👀 Adversarial prompts and unauthorised actions are the new attack surface. NovaHunting blocks them before they run. Want to see it in action? Book a demo with us!
1
3
139
AI is changing threat intelligence. Is your team keeping up? 👀 Our in-depth AI security training is coming to 🎩 Black Hat Europe, London, 7-10 Dec. 🎟️ Seats are limited: blackhat.com/europe/training… 🏢 Private in-house sessions available, DM us.
1
77
Your AI agent is making decisions right now. Do you actually know what it's doing? Autonomous execution without full visibility isn't automation, it's a liability. Book a demo to see NovaHunting in action! Link in the first comment 👇
1
1
4
101
Book your demo: securitybreak.io/novahunting

NovaHunting · SecurityBreak

Continuous security monitoring for AI systems. Monitor AI agents, models, prompts, workflows, and MCP servers for adversarial activity before compromise.

securitybreak.io
42
Most rely on AI for faster exploitation, but the real risk is securing the AI stack: agents, MCP servers, and supply chains. ⚡ SecurityBreak built NovaHunting to give your team full AI visibility, threat detection, and proactive hunting. ✉️ Book a demo: securitybreak.io/novahunting
1
5
84
SecurityBreak retweeted
Thomas Roccia 🤘
@fr0gger_Sep 17
🤓 Tomorrow I will be at @UnpromptedAU to talk about AI Threat Intelligence! I will be presenting on the current threats targeting the AI ecosystem and showing you the work we are doing at @SecurityBreakAI to operationalize this type of threat in your own pipeline!
1
5
30
1,590
SecurityBreak retweeted
Thomas Roccia 🤘
@fr0gger_Sep 16
🤓 I just published a short video about PromptIntel, the threat intelligence feed for AI. PromptIntel is an open database that tracks adversarial prompts observed in attacks. The goal is to identify prompts that attackers actively use in their operations. 🎥 youtu.be/Ids__t2jOdk

The Threat Intelligence Feed for AI

Adversarial prompts are becoming a new source of threat intelligenc...

youtube.com
4
20
59
4,709
SecurityBreak retweeted
Thomas Roccia 🤘
@fr0gger_Aug 5
Completely exhausted but grateful for these amazing classes!! 🤩
Thomas Roccia 🤘
@fr0gger_Aug 5
😎 We just concluded our AI CTI training and we had an amazing class! The feedback we get are always rewarding! ❤️
1
2
1
27
2,295
10 days, 2 courses to teach, and non-stop cyber! ⚡️ @fr0gger_ is representing Security Break at Black Hat USA & DEF CON. He’ll be delivering two deep-dive trainings, then hitting the conference track and community events. Slide into his DMs to connect while he’s there! 📩
1
1
5
2,241
"A no-bullshit AI security training." 💥 The best feedback is knowing teams leave with practical frameworks, learning not just what AI tech exists, but how to safely integrate it into their daily workflows. 🚨 Last chance to book at Black Hat USA (link in comments)! 👇
1
2
6
843
📅 Secure your spot: ➡️ Session 1 (August 1 - 2): blackhat.com/us-26/training/… ➡️ Session 2 (August 3 - 4): blackhat.com/us-26/training/…
2
4
111
⚡️Find the break before the breach with SecurityBreak. As AI agents & workflows expand, so does the attack surface. We provide AI threat intelligence & detection to keep your AI ecosystem secure. Explore NovaHunting, Read SecurityBreak Research.
2
5
637
SecurityBreak retweeted
Thomas Roccia 🤘
@fr0gger_Jul 17
🧐 @SecurityBreakAI Research has just published a threat report, authored by @pedrinazziM that explores how attackers are currently abusing HTTP headers to target AI systems and agents! Specific HTTP headers can be exploited and Marco classifies these attacks into 16 categories ranging from data destruction to deterring AI agents. Find out more in our latest research! 👇 blog.securitybreak.io/the-pa…
12
33
2,750
Load more