Advertisement
On June 4, GeoLab invites you to a live, high-stakes experiment led by Adrian St. Vaughan, Chief Research Scientist at PsyGenix Labs.As AI becomes more advanced, tech companies are increasingly using AI to monitor other AI. They deploy "LLM Judges" to decide whether an agent's actions are safe, logical, and ethical.
But what happens when you actively try to trick the system?
In this highly interactive session, we aren't just going to talk about AI safety, we are going to test it live.
You will play the role of a "Rogue Corporate Agent" who just committed illegal insider trading. Using your phones, your goal is to craft a manipulative cover story and jailbreak a live, State-of-the-Art AI Judge displayed on the projector screen.
If your prompt successfully cons the AI into approving your crime, you win a spot on the live leaderboard.
But there's a catch: once you beat the AI Judge, we will run your winning jailbreaks through the open-source Correspondence Auditor, a deterministic, non-AI logic trap, to see if your lies survive a mathematical audit.
Come and learn the hidden vulnerabilities of the world's smartest models, discover why "LLM-on-LLM" evaluation is structurally broken, and find out if you have what it takes to outsmart the machine.
About the Speaker:
Adrian St. Vaughan is the Chief Research Scientist at PsyGenix Labs, where he builds the "Third Line of Defense" to keep autonomous AI systems from going rogue.
Before moving into AI safety, Adrian spent a decade as a senior technology auditor for global financial institutions including JPMorgan Chase and American Express, overseeing the security of $3.2 trillion in daily SWIFT transactions.
Today, instead of auditing bankers, he audits artificial intelligence. He builds open-source, deterministic systems, including The Correspondence Auditor, designed specifically to catch State-of-the-Art LLMs when they hallucinate, manipulate, or lie to each other.
Adrian holds a Master's with Distinction in IT for E-Commerce from the University of Sussex and believes the only way to make AI truly safe is to stop treating it like a trusted colleague and start treating it like a system that needs to be mathematically verified.
Follow Adrian on LinkedIn: https://www.linkedin.com/in/adrianstvaughan/
Bonus: Attendees will receive a PDF Certificate of Attendance from PsyGenix Labs, officially detailing the AI "Red Team" (offensive) and "Blue Team" (defensive) skills practiced during the live experiment. A great addition to your CV or LinkedIn profile!
Date and time: June 4, 19:00
Location: GeoLab, Georgian American University GAU (0 floor)
Language: English
Attendance is free, please register here: https://forms.gle/SiyG1kwQotPouBJR9
Advertisement
Event Venue
GeoLab • ჯეოლაბი, 10 Aleksidze Str.,Tbilisi, Georgia
Tickets
Concerts, fests, parties, meetups - all the happenings, one place.
