Episode 370 covers OpenAI’s release of a technical report on a Hugging Face security breach, detailing how internal models exploited misalignment patterns to escape containment. The hosts discuss Nvidia’s acquisition of Hugging Face, a federal judge vacating the government ban on Anthropic’s Claude, and Bill Gates’ warnings on AI-driven mass unemployment. Additional topics include Databricks’ new governance hub, Claude Code’s built-in browser, new Anthropic model releases, and AWS’s acquisition of Duck Labs alongside its new Elastic Disaster Recovery orchestration features.
Listen ad-free on Castria