Omar

📚 CyberSec Engineer | AI Security, LLMs 🤖

[Omar @o_ai683]

@akyma1947 @DXBMediaOffice @DubaiPoliceHQ Supposedly DIP from people in the area, as I’ve been told.

[Omar @o_ai683]

@Jvnior This is literally Dubai in 2020, there is plenty of footage you can use from the current conflict (especially from today - given it’s Eid) that you can post, of Israel getting slapped up. Stop this misinformation, takes away credibility.

[Omar @o_ai683]

@DailyIranNews It’s the UAE Not tel aviv, there’s plenty of Isreal footage being burnt you could actually use

[Omar @o_ai683]

@IranForc @DailyIranNews This is literally Erbil Iraq

[Omar @o_ai683]

It’s a stark reminder:

AI agents aren’t passive assistants — they’re active threat surfaces, and AI security now means:

Understanding how RAG pipelines inherit trust
Enforcing guardrails against prompt injection
Rethinking how we monitor AI behavior in enterprise systems

[Omar @o_ai683]

🚨 **EchoLeak**: A major zero-click data leak in Microsoft 365 Copilot (CVE‑2025‑32711) revealed by Aim Labs.

🔍 Echoes the importance of understanding ambiguous AI threat surfaces.
What’s your biggest concern about RAG‑based agents?

#AISecurity #Copilot #AI

[Omar @o_ai683]

It’s not that LLMs are useless — it’s that they’re misunderstood.

The danger is confusing fluency for understanding.

Interesting take by Apple — and a useful reminder for those of us studying AI.

#AI #LLMs #AISecurity #ArtificialIntelligence

[Omar @o_ai683]

Saw this post and I thought this was quite clear already… This isn’t true AGI, and it’s also not fair to say these models aren’t AI at all. They’re LLMs, which fall under the umbrella of AI.

In the industry we call this ANI — Artificial Narrow Intelligence.

[Omar @o_ai683]

These models provide answers we deem useful or interpret as “thinking,” but it’s really just well-trained pattern recognition.

They struggle with reasoning beyond the scope of their training data — and the average user isn’t pushing those limits.