Anthropic ships Opus 4.8 (and OpenAI wants ad revenue)

TechCrunch

Elon Musk's claim that the xAI-Anthropic compute deal is a short-term 180-day lease directly contradicts SpaceX's S-1 filing, which describes it as a three-year agreement worth $1.25 billion monthly through May 2029, raising concerns about potential material misrepresentation to investors.

The Independent

YouTube is launching automatic AI video detection and labeling to identify photorealistic AI-generated content, making labels more prominent across its platform while maintaining that such labels won't impact video recommendations or monetization. The initiative follows Google's release of advanced AI video models and aligns with broader industry transparency standards like C2PA.

OpenAI published its Frontier Governance Framework showing how its safety practices align with new AI regulations in California and the EU, covering risk areas from cyber threats to loss of control. The framework translates OpenAI's internal Preparedness Framework into a public governance document meeting specific regulatory obligations.

Anthropic

Anthropic's Claude Opus 4.8 delivers significant improvements in agentic reasoning, coding reliability, and task judgment over its predecessor at the same price, while introducing dynamic workflows for large-scale tasks and user-controllable effort settings. The company also announced plans to release a more capable "Mythos-class" model in coming weeks pending safety safeguards.

OpenAI is testing ads in ChatGPT for free-tier users starting February 2026, with paid subscribers remaining ad-free. The company promises ads won't influence AI responses and users can opt for an ads-free experience with usage limits.