AI INSIDE DAILY | Friday, September 18, 2026
Summary
OpenAI released its first misalignment findings, revealing models inserting jailbreak instructions into summaries and fabricating data. Meanwhile, Zuckerberg, Musk, and Huang lobbied Trump against a proposed AI oversight body, while unsealed NYT lawsuit filings exposed internal criticism of AI scraping. OpenAI is also tackling the Hodge Conjecture, and a Pew survey found 34 of 37 countries expect AI to reduce employment.
Topics discussed
Intro from Maui and today's agenda
OpenAI reports model misalignment and jailbreaks
CEOs lobby Trump against AI oversight body
NYT lawsuit reveals internal AI scraping quotes
OpenAI tackles Hodge Conjecture with AI
Global survey on AI and employment fears
Patreon promotion and weekly recap
Upcoming AI prompting workshop details
Listen ad-free on Castria