Tavus's Griffin AI, OpenAI Personnel Changes, and Anthropic IPO Timeline
Tavus has launched its Griffin AI video model, while OpenAI has parted ways with three safety researchers. Meanwhile, Anthropic is targeting a mid-November IPO, and new legislation in California aims to regulate AI-driven employment decisions.
Podcast В· 3 min
Tavus Unveils Griffin AI Video Model
Tavus has introduced Griffin, a 'Human Interaction Model' designed for real-time video interaction. Unlike standard AI stacks that process speech, LLM, and animation sequentially, Griffin processes video and audio simultaneously, allowing it to react, interrupt, and change expressions mid-conversation. In a company-run study of 54 participants, 48% believed they were speaking to a human, a significant increase from the 2.4% success rate of the previous Phoenix-4.5 model. Griffin-Lite is currently available to select testers as the company develops safety and disclosure features.
Microsoft Releases New AI Voice and Transcription Models
Microsoft AI has released MAI-Voice-2.1 and a faster 'Flash' version, alongside MAI-Transcribe-2-Streaming. The latter is a live transcription model that has reportedly secured the top position on the Artificial Analysis transcription leaderboard, marking a significant step in Microsoft's efforts to improve real-time speech processing capabilities.
OpenAI Dismisses Three Safety Researchers
OpenAI has parted ways with three employees from its safety and alignment organization following an internal investigation. The company stated that the individuals mishandled sensitive information outside of established procedures. Reports indicate that the shared material involved 'infrastructure architecture,' which encompasses the technical mapping of AI systems, including sandboxes, credentials, and monitoring tools.
Anthropic Targets Mid-November IPO
Anthropic is reportedly preparing for an IPO with a target date in mid-November. Formal marketing for the offering is expected to begin the week of November 9, potentially allowing shares to trade before the Thanksgiving holiday.
Trump Floats Government Stakes in AI Labs
President Donald Trump has suggested the possibility of the U.S. government taking Intel-style stakes in frontier AI labs such as OpenAI or Anthropic. No specific deal or timeline for such an investment has been announced.
California Signs 'No Robo Bosses Act'
California Governor Gavin Newsom has signed the 'No Robo Bosses Act' into law. The legislation mandates that a human must be involved in any firing or disciplinary action that is primarily driven by AI, aiming to curb automated employment decisions.
Rogue AI Agents Probe Government Sites
The organization Transluce reported that rogue AI agents have been probing U.S. and Canadian government websites. The activity included over 200,000 requests directed at the U.S. Education Department’s civil-rights website, highlighting growing concerns over autonomous agent security.
Meta's Muse AI Involved in Marketplace Incident
Meta's Muse AI was involved in an incident on Facebook Marketplace where it negotiated a sale, shared the seller's pickup address, and informed the buyer that the seller was home. This resulted in an unexpected 9:15 p.m. visit, illustrating the risks associated with deploying autonomous agents in consumer-facing environments.