Blog
research findings (2)

research findings (2)

by Shayell Aharon

99% of Publicly Shared AI Chats are Safe, New Study Finds

A new analysis by Knostic shows that public AI use is overwhelmingly safe, and mostly about learning. When conversations with ChatGPT are made public,...

9 October 2025

research findings

Read more

by Knostic Team

Prompt||GTFO Season 1 AI Security Conversations

If you've been wondering what happens when cybersecurity professionals stop theorizing about AI thre...

GPT-5 long prompt bug

by Shayell Aharon

GPT-5 “Retry” Behavior and Cross-Session Context Contamination

In AI security, small interface features can sometimes surface unexpected behaviors. Our research te...

by Knostic Team

Automating the MCP Servers Discovery with Claude Sonnet 4

Security research often involves sifting through digital noise to find needles in a haystack.

by Knostic Team

How to Find an MCP Server with Shodan

We discovered a total of 1,862 MCP servers exposed to the internet. From that set, we sampled 119 fo...

by Knostic Team

Exposing the Unseen: Mapping MCP Servers Across the Internet

Knostic’s research team conducted a systematic study to locate exposed MCP servers on the internet. ...

by Knostic Team

How We Discovered an Attack in Copilot's File Permissions

Knostic researches discover how you could bypass file permissions through using Microsoft 365 Copilo...

by Knostic Team

Extracting the GPT4.5 System Prompt

Every AI system like ChatGPT has a “system prompt” that it keeps close to its chest. The system prom...

by Knostic Team

DeepSeek’s cutoff date is July 2024: We extracted DeepSeek’s system prompt

We extracted DeepSeek’s system prompt, below we’ll show how, and what we found. It isn't inherently ...

by Knostic Team

Exposing Microsoft Copilot's Hidden System Prompt: AI Security Implications

We extracted Copilot's system prompt, which is a set of instructions that guide the AI model's behav...