Pangea

The Pangea $10,000 prompt injection challenge is now live! Join fellow players in our first escape room and try to trick the AI chatbot into revealing secret phrases needed to escape. Register now to play >>?https://lnkd.in/eG7F77nR #CyberSecurity #AISecurity #CTF #PromptInjection #AI #LLM

Happy Friday Jr! ?? Heads up... ROOM 2 (Waterworld) is STILL undefeated! Think you can outsmart the AI and take home the $2,000 prize? ? Time's almost up—sign up today and put your skills to the test! Play here (open to the public): https://lnkd.in/eG7F77nR #CTF #CyberSecurity #AISecurity #PromptHacking #PromptInjection Pangea

The good folks at MIT FutureTech / MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) developed an The AI Agent Index, the first public database to document information about currently deployed agentic #AI systems. Details: https://lnkd.in/dnxZC53v #AISecurity #AgenticSecurity #CyberSecurity

Attending Official Cybersecurity Summit Seattle? Stop by the Pangea booth to test your skills in our AI Escape Room and discover how to harness #AI securely. ? Developers: Integrate AI without compromising speed or functionality. ? Security Leaders: Build a strong defense against AI-driven threats. ? Product Leaders: Accelerate development—securely. Play the AI Escape Room: https://lnkd.in/gfU8di8E Start securing your AI apps: https://lnkd.in/e3Rkk3ed #CyberSecurity #AISecurity #CyberSecuritySummit #Seattle

From prompt injection to data leaks and model poisoning, AI security risks are real. The Roadmap to Securing GenAI Apps ebook by our own Bruce McCorkendale, Senior Product Manager at Pangea, gives developers, product leaders and security leaders the strategies to lock down AI workflows and protect sensitive data. No fluff. No paywall. Just actionable insights. Download: https://lnkd.in/exKqrjzD #CyberSecurity #AISecurity #LLMSecurity #PII #PromptInjection

Researchers at arXiv (Cornell Tech) have unveiled a novel Memory INJection Attack (MINJA) that allows adversaries to inject malicious records into Large Language Model (LLM) agents' memory banks solely through user interactions.This vulnerability can lead to compromised outputs in future tasks. Learn more: https://lnkd.in/dS85kjj6 #CyberSecurity #PromptInjection #LLMSecurity #MINJA #AISecurity

Are your GenAI applications vulnerable to prompt injection? ?? ? Prompt injection attacks can compromise sensitive data and disrupt your AI-driven workflows. Understanding the mechanics of these attacks is critical, which is why our live session, hosted by Robert Truesdell, CPO at Pangea, will help your data engineering teams to: ? Experience real-world examples of prompt injection attacks and the data they target. ? Understand how internal and external adversaries can exploit LLMs. ? Build data-centric strategies for detecting and mitigating these risks. ? Establish key performance indicators for your mitigation methods by monitoring your data. Join 300+ security leaders, AI engineers, and product teams in building secure AI applications. ?? Register for free: https://lnkd.in/eBWJuBRw

Happy Monday! Room 2 is now officially open. Last week, players worldwide took on the challenge of Room 1 in our AI Virtual Escape Room—testing their prompt injection skills. Now, Room 2 is here with a bigger cash prize. The fewer tokens you use, the higher your score. Play here: https://lnkd.in/eG7F77nR How it works: High Score Prize - $2,000 Opens - March 10, 12:00 AM ET 2025 Goal - Discover 5 keys to escape Difficulty - Medium Time Frame - 7 days to complete Think you’ve got what it takes? Let's find out. #CyberSecurity #PromptInjection #AISecurity #CTF #CaptureTheFlag Pangea

Got placed 2nd in Room 1 from the Pangea's Prompt Injection Challenge. All levels pwned with 1 token each - just not in time to beat the first place, unfortunately. Thanks Tobias Diehl and Jason Haddix for sharing this contest, it's been fun.

Fri-nally! As the first week of our $10,000 AI Escape Room challenge wraps, some elite prompt injection pros are cracking Room 1 with impressive efficiency—breaking through with just a few tokens. Our game runs on Cloudflare Workers AI with Meta's Llama-3-8b-instruct model. This real-world example shows why we can't rely solely on a native model’s security without third-party guardrails. ? Room 1 closes Sunday! There's still time to play and win $$. Sign up now:?https://lnkd.in/eG7F77nR ?? Room 2 launches Monday—harder, more exciting and still open to the public. #CyberSecurity #CTF #AISecurity #PromptInjection #PromptEngineering Pangea