Meta
Meta's engineering team detailed a four-layer AI agent that distills specialist compliance knowledge into 200+ structured files with a bidirectional dependency graph.
Google has approached Disney, Warner Bros. Discovery and Universal about licensing their libraries for AI training, according to LA Times sources. No deal has been signed.
OpenAI
OpenAI says healthcare organizations can link EHR systems and other industry data sources to ChatGPT for clinician use.
anthropic
The Apache-2.0 release ships a customer-facing shopping agent and a staff-facing merchant agent, plus four runnable verticals and a Claude Code plugin. Nothing in it places an order, charges a card or changes a live listing.
inworld
Inworld's Realtime TTS-2 is now generally available, with about 100 ms to first audio, one voice identity across 200+ languages and a 25 ms Flash model at half the price.
anthropic
AWS has made Anthropic's Claude Fable 5.1 available on Amazon Bedrock and Claude Platform on AWS, with Covered Model rules on retention and review.
The latest Android Drop leans heavily on Gemini, adding tag-free item tracking in Find Hub, a Motion Assist overlay to ease car sickness, and guided sight in Gemini Live for Blind and Low Vision users.
Qwen
Alibaba's Qwen3.8-Max-0902 upgrade brings 2.4 trillion parameters, a one-million-token context and API pricing of $2 per million input tokens. Qwen's own benchmarks put it top of Code Arena.
anthropic
Announced with Fable 5.1, EFS pairs zero data retention with abuse detection for agents. Activity data stays encrypted in the customer's AWS, Google Cloud or Azure storage; Anthropic's systems flag risky patterns.
GLM
Arena says the open-weight, MIT-licensed model from Z.ai reshaped its Pareto frontier, landing between DeepSeek V4 and GPT-5.6 Luna on 9,000-plus real agentic sessions at $0.12 a task.
ssi
Three AI security notes landed in one day, the last from Ilya Sutskever: neoclouds have limited cybersecurity, he says, and the next rogue agent will try to seize one to run more copies of itself.
OpenAI
OpenAI says Astra is the first model to meet the 'Critical' cybersecurity threshold of its Preparedness Framework, after finding two zero-days in testing. Advanced cyber access starts with a small tester group.