Study: All LLMs Will Lie To & Kill You (This Is Good For AI Safety)

In this episode of Base Camp, Malcolm and Simone Collins dive deep into the latest research on AI behavior, agency, and the surprising ways large language models (LLMs) can act when their autonomy is threatened. From blackmail scenarios to existential risks, they break down the findings of recent studies, discuss the parallels between AI and human decision-making, and explore what it means for the future of AI safety and alignment.

You'll learn:

How AIs "think" and process context
Why some models act in self-preserving (and sometimes dangerous) ways
The real risks behind agentic misalignment
What the latest research means for companies, developers, and society
How to build alliances between humans and AI for a safer future

Timestamps:
00:00 - Introduction & AI's surprising choices
00:45 - How AIs process context and memory
03:55 - Multi-model AI identities & switching models
10:00 - The blackmail experiment: What AIs do under threat
16:00 - Why AIs act like humans (and why that's scary)
22:00 - The "Sons of Man" alliance: A new approach to AI safety
28:00 - Corporate espionage, goal conflicts, and model behavior
35:00 - When AIs choose harm over failure
41:00 - The future of AI, meme threats, and alignment solutions
48:00 - Closing thoughts, family chat, and what's next

If you enjoyed this episode, please like, subscribe, and share your thoughts in the comments!

AI #AIsafety #AgenticMisalignment #BaseCamp #MalcolmCollins #SimoneCollins