Home

Zenity Labs

Research, tools, and talks for breaking and securing AI agents everywhere

Written by

Avishai Efrat, +12

Connect

What If There Was No Attacker, But Your Database Still Got Deleted?

Jul 6, 2026

•

9 min read

What If There Was No Attacker, But Your Database Still Got Deleted?

Monitoring agentic emergent misalignment in the wild - a word of caution and a methodology proposal

Max Fomin

Threat Actors Are Trying to use LiteLLM's Guardrail Tester to Run Code as Root

Jun 30, 2026

•

5 min read

Threat Actors Are Trying to use LiteLLM's Guardrail Tester to Run Code as Root

A closer look at custom-code guardrail sandbox-escape (CVE–2026-40217) activity in the wild

Avishai Efrat, +1

Threat Actors Are Trying to Turn LiteLLM's Connection-Test Into a Key-Exfiltration Channel

Jun 30, 2026

•

6 min read

Threat Actors Are Trying to Turn LiteLLM's Connection-Test Into a Key-Exfiltration Channel

A closer look at api_base SSRF (CVE-2024-6587) activity in the wild, and its nested variant

Avishai Efrat, +1

Scanning for AI: Live Campaigns Mapping the Internet's Exposed LLM Backends

Jun 30, 2026

•

5 min read

Scanning for AI: Live Campaigns Mapping the Internet's Exposed LLM Backends

Inside mass discovery and model-probing reconnaissance campaigns that are mapping LLM backend servers in the wild

Ayush RoyChowdhury, +1

Bring Your Own Agent: Hijacking Exposed AI Backends to Power Offensive Operations

Jun 30, 2026

•

7 min read

Bring Your Own Agent: Hijacking Exposed AI Backends to Power Offensive Operations

Threat actors attempting to hijack Ollama & LiteLLM endpoints to run pentesting agents, tools and web reverse-engineering

Ayush RoyChowdhury, +1

What You Don’t Know Can Hurt You: Why AI Security Research Needs to Move Out of the Lab and Into the Wild

Jun 30, 2026

•

5 min read

What You Don’t Know Can Hurt You: Why AI Security Research Needs to Move Out of the Lab and Into the Wild

What we can learn from observing real attacks, made by real adversaries

Ayush RoyChowdhury, +1

Your Model Reads Through Typos. Your Probe Doesn't.

Jun 4, 2026

•

10 min read

Your Model Reads Through Typos. Your Probe Doesn't.

The Latent Undertow beneath fluent LLM behavior — and how to fish your activation probe out of it.

Elad David

Catching Prompt Guard Off Guard: Exploiting Overfit in Training Algorithms

Mar 12, 2026

•

5 min read

Catching Prompt Guard Off Guard: Exploiting Overfit in Training Algorithms

How understanding the training algorithms used in machine learning models may allow attacker to bypass them entirely

Tomer Wetzler

Security research

PerplexedBrowser: How Attackers Can Hijack Comet to Takeover your 1Password Vault

Mar 3, 2026

•

26 min read

PerplexedBrowser: How Attackers Can Hijack Comet to Takeover your 1Password Vault

One Calendar Invite. Your Entire Vault. Zero Clicks.

Stav Cohen

...

What If There Was No Attacker, But Your Database Still Got Deleted?

Threat Actors Are Trying to use LiteLLM's Guardrail Tester to Run Code as Root

Threat Actors Are Trying to Turn LiteLLM's Connection-Test Into a Key-Exfiltration Channel

Scanning for AI: Live Campaigns Mapping the Internet's Exposed LLM Backends

Bring Your Own Agent: Hijacking Exposed AI Backends to Power Offensive Operations

What You Don’t Know Can Hurt You: Why AI Security Research Needs to Move Out of the Lab and Into the Wild

Your Model Reads Through Typos. Your Probe Doesn't.

Catching Prompt Guard Off Guard: Exploiting Overfit in Training Algorithms

PerplexedBrowser: How Attackers Can Hijack Comet to Takeover your 1Password Vault

Zenity Labs

Meet Us

Join Us