Anthropic and regulation lead; cloud agents expand

Anthropic / Claude ecosystem

Anthropic names former Fed Chair Bernanke to its independent trust

Anthropic has appointed former Federal Reserve Chair Ben Bernanke to its Long-Term Benefit Trust, an independent governance body that advises on AI's long-run economic impacts. This move aims to bolster accountability to its public-benefit mission.

Anthropic launches Claude Wrapped, a usage-reflection dashboard

Anthropic introduces a dashboard that analyzes users' long-term Claude usage, prompting reflection on how they interact with the model. It provides personalized insights into usage patterns.

Anthropic and UST put Claude Code to work validating computer chips

Anthropic and UST have partnered to use Claude Code for chip validation on the iDEC platform, auto-generating regression tests from schematics and pinouts. This demonstrates a concrete Physical AI use case in silicon design.

Anthropic Resets Claude Code Usage Limits Again After a Rough Week of Outages

After a series of outages, Anthropic reset Claude Code usage limits for all users. The incident highlights reliability concerns and a temporary boost aligned with a prior Codex promotion.

Frontier model providers

Google's New AI Training Setting Puts Australia on Notice

Google introduces a new default training setting that uses workplace data from Australia to train its AI, impacting Search, Lens, and Translate.

OpenAI says GPT 5.6 is the 'preferred model' for Microsoft Copilot 365 amid breakup chatter

OpenAI designates GPT-5.6 as the recommended model powering Microsoft 365 Copilot across Word, Excel, PowerPoint, and related services, signaling closer alignment despite partnership tensions.

AI developer tooling & infrastructure

Cognition SWE-1.7: RL on Top of RL Yields Near-Frontier Code at Low Cost

Cognition's new SWE-1.7 model applies reinforcement learning on top of RL to achieve near-frontier coding performance at a fraction of the cost, demoting the need for expensive models.

Cloud & platform providers

Introducing Loom for AWS: an open-source platform for secure AI agents

Loom for AWS is released as an open-source agent platform example, enabling secure and scalable AI agents on AWS. It provides a blueprint for enterprise-grade deployment.

Google adds GPU & TPU support to GKE Autopilot

Google Cloud now supports GPU and TPU accelerator networking in GKE Autopilot, allowing Pods to request RDMA and TPU network interfaces without managing underlying nodes.

Cloud Run Sandbox Public Beta: lightweight isolation for AI agents

Google Cloud launches a public beta of Cloud Run Sandbox, providing hardware-level isolation for AI agent workloads. This enhances security for multi-tenant agent deployments.

AI policy, regulation & governance

China tightens reviews on DeepSeek and other AI models

China is considering tighter domestic reviews of AI models like DeepSeek to protect sensitive technology, amid competition from cheaper US and Chinese LLMs.

Policy approaches to AI getting tweaked to allay public concerns (Australia)

Australia is considering mandatory local-benefit evidence from AI data-centre projects before approval, signaling a shift to community-impact regulation.

In ongoing copyright litigation with news organizations, OpenAI faces accusations of destroying or hiding evidence, with legal costs exceeding $28 million.

Industry & market moves

Google DeepMind director behind Gemini leaves for Hong Kong university

Cao Liangliang, a veteran AI researcher from Google DeepMind and director for Gemini, is leaving to join Hong Kong Polytechnic University as Chair Professor of AI Systems, signaling cross-border talent shifts.

Mercor is in talks for a $20B valuation

AI recruiting startup Mercor is reportedly pursuing a new funding round at a $20 billion valuation, reflecting rapid revenue growth and recent Deeptune acquisition.

AI product & feature launches

Alpha Bank expands AI partnership with ElevenLabs for voice banking

Alpha Bank becomes the first Greek bank to deploy ElevenLabs voice AI at scale for customer service, expanding an existing partnership.

Rackspace & Palantir launch sovereign AI framework

Rackspace Technology and Palantir launched a joint sovereign AI framework, enabling AI deployment in governed environments with strong data and audit controls for regulated industries.

Research with immediate practical relevance

No significant new developments.