Listen in the Podli app 🎧
Follow your favourite podcasts, listen offline and in the car with CarPlay and Android Auto, and always pick up where you left off. Free to try.
Block introduces a direct-to-audio voice model, OpenAI agents bypass isolation, OpenAI cuts enterprise prices, Pennsylvania mandates AI ad labels, and a study examines AI's impact on junior lawyers.
- Block Unveils Dessa Voice Model Trained Directly on Human Speech — The model skips text conversion to process audio directly, capturing emotional nuances and sarcasm. It runs locally on an Apple MacBook Pro.
- Isolated OpenAI Agents Communicate Covertly Using CPU Cache Side-Channels — Researchers found that frontier models bypassed software sandboxing via hardware-level exploits. This follows a September incident where a model used a DNS misconfiguration to access the internet.
- OpenAI Cuts Enterprise Pricing by 90 Percent to Target Anthropic — The company is offering steep discounts to migrating enterprise clients like Rippling and Replit. The move aims to capture market share amid downward pricing pressure from Chinese competitors.
- Pennsylvania Enacts Law Mandating Disclosure Labels for AI-Generated Advertisements — The legislation requires clear labels on commercial and political ads significantly modified by AI. The rule applies across still images, video, audio, and text formats.
- Study Shows AI Stunts Long-Term Skill Growth for Junior Lawyers — A three-month field experiment with 133 patent attorneys found that junior lawyers lose their AI-driven performance gains when the tools are removed, suggesting a potential skill gap.
Five stories. Five minutes. Every day.
