Faouzi EL YAGOUBI

Building trustworthy AI for real-world impact. AI researcher, serial entrepreneur & YC founder.

[Faouzi EL YAGOUBI @yagobski]

Excited to be speaking at IWPE’26, co-located with IEEE EuroS&P in Lisbon, where I’ll present our paper: “The Interlocutor Effect: Why LLMs Leak More Privacy to Agents Than Humans.”
https://t.co/BKyPTY5qWW
📅 July 6, 2026

[Faouzi EL YAGOUBI @yagobski]

Excited to share that our paper, “AgentLeak: A Benchmark for Internal-Channel Privacy Leakage in Multi-Agent LLM Systems,” has been accepted in IEEE Access!

We show why privacy audits must look beyond final outputs and inspect internal agent channels.

🔗 https://t.co/o7W03fXCjZ

[Faouzi EL YAGOUBI @yagobski]

I just published The Interlocutor Effect: Why LLMs May Leak More Personal Data to AI Agents Than to Humans https://t.co/IaxlgzGNdw

[Faouzi EL YAGOUBI @yagobski]

I just published AgentLeak: Why Privacy Audits Must Look Inside Multi-Agent LLM Systems https://t.co/MG7cVCWmyq

[Faouzi EL YAGOUBI @yagobski]

Do LLMs trust bots more than humans? 🤖 > 🧑‍🦱
"The Interlocutor Effect: Why LLMs Leak More Personal Data to Agents Than Humans". We found that AI privacy guardrails fail more often depending on who the AI thinks it's talking to!
https://t.co/ZxPvq3rI1L

[Faouzi EL YAGOUBI @yagobski]

I am excited to announce that I will be presenting our work on the Interlocutor Effect at #IWPE in Lisbon! 🇵🇹🗣️
We dive into a critical AI privacy question: do LLMs leak more sensitive data depending on who they think is prompting them? 🕵️‍♂️🔒

https://t.co/BKyPTY5qWW

[Faouzi EL YAGOUBI @yagobski]

We tested GPT-4o, Claude-3.5-Sonnet, Llama-3.3-70B, Mistral-Large across 1,000 scenarios (healthcare, finance, legal, corporate).

32-class attack taxonomy. 4 frameworks: LangChain, CrewAI, AutoGPT, MetaGPT.

Full paper: https://t.co/GTmiGtrUOs

[Faouzi EL YAGOUBI @yagobski]

THREAD: We built AgentLeak — the first full-stack benchmark for privacy leakage in multi-agent LLM systems.

Here's what surprised us.

[Faouzi EL YAGOUBI @yagobski]

Counterintuitive result: multi-agent setups actually REDUCE per-channel output leakage vs single-agent (27.2% vs 43.2%).

But the attack surface doesn't shrink. It moves to channels nobody's watching.

That's what makes it dangerous.

[Faouzi EL YAGOUBI @yagobski]

New report: 48.9% of organizations are completely blind to machine-to-machine traffic from AI agents.

We're deploying autonomous systems at enterprise scale while 92% lack the security maturity to defend them.

This isn't a future risk. It's the current state.

[Faouzi EL YAGOUBI @yagobski]

"The dark side of autonomous intelligence: a survey on data leakage and privacy failures in agentic AI" — Read it on @ResearchGate: https://t.co/4zEnJ5tfRF

[Faouzi EL YAGOUBI @yagobski]

AgentLeak — privacy benchmark for multi-agent LLMs

→ 68.8% of leakage via inter-agent msgs (invisible to output audits)
→ 41.7% missed by standard evaluation
→ 7 channels tracked across AutoGen/LangGraph/CrewAI

https://t.co/GTmiGtrUOs

#LLM #AIPrivacy #MultiAgent

[Faouzi EL YAGOUBI @yagobski]

The "Open Source is dead" crowd has it backwards.

In the AI era, Open Source + Self-Improving Software = a living asset.

Why buy a rigid tool when you can deploy a system that watches how you work and patches itself to fit your needs? No APIs required.

[Faouzi EL YAGOUBI @yagobski]

Excited to share our latest research: "AgentLeak: A Benchmark for Evaluating Privacy Leaks in LLM Multi-Agent Systems".
https://t.co/hdwl9rPp8a

[Kim Dotcom @KimDotcom]

Sharing this nuclear war video is the most important thing you can do to prevent nuclear war. Thank you �

[Faouzi EL YAGOUBI @yagobski]

Like what we do? Leave us a review and help us spread the word! https://t.co/pNtNhWjHPC

[Elon Musk @elonmusk]

🇲🇦🇲🇦 Congrats Morocco!! 🇲🇦🇲🇦

[Faouzi EL YAGOUBI @yagobski]

@tog3pies @MusicHater75 @Rogers Are they are using Rogers network to response?

[Faouzi EL YAGOUBI @yagobski]

@elonmusk What do you think to be the first ?