Frontier labs ship voice, image models; Illinois AI law

Anthropic / Claude ecosystem

Anthropic J-lens Reveals Hidden Workspace Inside Claude

Anthropic researchers discovered a 'J-space' global workspace inside Claude models, accessible via a new interpretability tool called J-lens. The study shows that verbalizable representations form a workspace that can influence outputs and reveals private recognition of testing scenarios.

Claude Cowork expands to mobile and web

Anthropic's Claude Cowork, an AI agent similar to Claude Code, is expanding from desktop to web and mobile platforms for Max subscribers, enabling task pickups across devices.

Anthropic eyes massive Australia data centre grab

Anthropic is reportedly seeking up to ~1.4 GW of Australian data center capacity by 2027 to support its AI infrastructure and potential IPO plans.

Frontier model providers

OpenAI Releases GPT-Realtime-2.1 and GPT-Realtime-2.1-mini for Low-Latency Voice Agents

OpenAI expanded its Realtime API with two new low-latency voice models, including a faster mini reasoning option, to support voice agent development.

Introducing Muse Image: Image Generation Built for Your World

Meta launched Muse Image, its first image generation model from Meta Superintelligence Labs, integrated across Meta AI, Instagram, and WhatsApp, with advertiser tools coming soon.

xAI Expands Grok Voice with 21 Multilingual AI Voices

SpaceXAI (formerly xAI) expanded Grok Voice with 21 new multilingual voices, enhancing its real-time speech platform for enterprise and developer use.

AI developer tooling & infrastructure

MCP Goes Stateless on July 28: Is Your Audit Trail Still Intact?

The upcoming MCP stateless release shifts audit-trail management from session IDs to application-owned evidence identities, requiring new host-issued identifiers.

Cloud & platform providers

Amazon SageMaker Studio Integrates Hugging Face: One‑Click AI Deployment

AWS added a one‑click deep‑link from Hugging Face to SageMaker Studio, enabling immediate model loading, customization, and deployment.

NVIDIA and Hugging Face Bring New Models and Frameworks to LeRobot for Open Robotics

NVIDIA and Hugging Face contributed new models and frameworks to the open-source LeRobot platform, advancing community-driven robotics development.

AI policy, regulation & governance

Illinois Enacts Frontier Model Safety Law

Illinois passed SB 315, a frontier model safety law mandating annual third‑party audits and reporting for large AI models, aligning with similar efforts in New York and California.

Industry & market moves

TeraWulf Announces Anthropic Lease at Justified Data Campus and Sale of Abernathy Joint Venture to Fluidstack

TeraWulf has entered into a long-term lease agreement with Anthropic at its Justified Data Campus, and separately sold its majority interest in the Abernathy joint venture to Fluidstack. The deals free capital for TeraWulf to pursue wholly owned AI infrastructure opportunities.

Google DeepMind Expands Gemini AI Team for Agent Post‑Training Research

Feryal Behbahani announced hiring for a new research role focused on post‑training for Gemini agents, signaling Google DeepMind's push toward more autonomous AI systems.

xAI Is Now Officially Known As SpaceXAI

Elon Musk's xAI has rebranded to SpaceXAI, consolidating AI operations under the SpaceX umbrella after its IPO, with a new logo and branding.

Honeywell spinoff Solstice to acquire Element Solutions in $14.5B deal

Solstice Advanced Materials, a Honeywell spinoff, agreed to acquire Element Solutions for $14.5 billion in a cash-and-stock deal to expand its AI infrastructure business.

Stardust AI Closes RMB 1B New Round Led by SASAC's Chengtong Fund

Chinese embodied AI startup Stardust AI (Xingdong Jiyuan) closed a RMB 1 billion Series C round, bringing its two‑month total to RMB 2.5 billion.

Ondas Announces Transformative Acquisition of DZYNE Technologies

Ondas Inc. acquired DZYNE Technologies for $875.8 million, establishing what it calls the leading autonomous defense platform spanning ISR, counter‑UAS, and autonomous effects.

Noon Academy Acquires AlMakhfy to Expand Saudi EdTech

Noon Academy acquired AlMakhfy to broaden its AI‑powered education offerings in Saudi Arabia.

US cyber agency is using Anthropic’s Mythos to audit government code

CISA is reportedly using Anthropic's Mythos, a private offensive‑grade AI model, to audit government code for vulnerabilities.

AI product & feature launches

Ray Kurzweil Unveils RAI – a Human‑First Digital Twin

Ray Kurzweil announced RAI, a new Human‑First Digital Twin based on his knowledge framework, as a joint venture debuting at the UN ITU AI for Good Global Summit.

Sber launches GigaChat 3.5 Ultra as open source AI

Russian bank Sber released GigaChat 3.5 Ultra as a free, open-source AI model for developers and users.

Robbyant Unveils LingBot-Depth 2.0 and LingBot-Vision to Redefine Robotic Spatial Perception

Robbyant launched LingBot‑Depth 2.0, a next‑gen depth perception model, and LingBot‑Vision, a visual foundation model, advancing robotic spatial understanding.

Research with immediate practical relevance

No significant new developments.