AI firms are going back on their safety promises
Summary
Marketplace Tech examines the Future of Life Institute’s latest report card, which reveals declining safety scores for major AI labs like Anthropic, OpenAI, and Google. Guest Sabina Nong highlights concerning trends, including companies weakening voluntary safety pledges and increasing real-world harms linked to their systems. The discussion critiques the competitive race toward superintelligence and advocates for stronger regulatory frameworks to ensure AI remains under human control.
Topics discussed
Sponsors: Cornerstone and Odoo
AI safety report cards show slipping grades
Companies weakening safety commitments and causing harms
The danger of recursive self-improvement and competitive racing
Sponsors: Cornerstone and Dell Pro
Critique of the race to superintelligence
Need for regulatory action and global safety standards
Teaser: Geoengineering solutions on How We Survive
Listen ad-free on Castria