EU AI Act takes effect; AI agents gain ground

Anthropic / Claude ecosystem

Anthropic and UST put Claude Code to work validating computer chips

Anthropic's Claude Code has been deployed in partnership with UST to automatically generate chip validation regression tests. This collaboration is reported to reduce validation cycle times by 50–70%.

Frontier model providers

China AI model developers ramp up in-house chips: DeepSeek, Zhipu explore custom silicon

Chinese AI developers like DeepSeek and Zhipu AI are pursuing custom silicon solutions to reduce their reliance on US-restricted NVIDIA chips. This initiative aims to deepen partnerships with domestic suppliers such as Huawei.

ChatGPT is coming for one of Google's smartest Chrome features

OpenAI has launched a ChatGPT Chrome extension that offers context-aware webpage understanding. This directly competes with Google's Gemini in Chrome browser integration.

AI developer tooling & infrastructure

Cursor Is Developing an AI Agent to Compete With Claude Cowork

Cursor, a code editor backed by Khosla Ventures, is developing its own AI agent designed to compete directly with Anthropic's Claude Cowork offering.

Cloud & platform providers

Building secure AI agents at scale: Introducing Loom for AWS

AWS has released Loom, an enterprise-grade agent platform designed for integrating governance, security, and lifecycle management of AI agents at scale on Bedrock AgentCore.

Google adds GPU & TPU support to GKE Autopilot

Google has extended GKE Autopilot to include native GPU and TPU networking support. This allows managed accelerator workloads without the need for manual node configuration.

Amazon EC2 I7ie instances now available in AWS Asia Pacific (Hyderabad) region

AWS has launched its I7ie storage-optimized compute instances in the Hyderabad region. These instances offer 40% better compute performance and 65% better storage I/O performance compared to the prior I3en generation.

Cloud Run Sandbox Public Beta: A New Lightweight Isolation Paradigm for the AI Agent Era

Google Cloud has launched a public beta of Cloud Run Sandbox, offering lightweight hardware-level isolation for AI agent workloads. This solution promises 70% cost savings compared to full VMs while maintaining strong security boundaries.

Introducing AWS Sandbox Environment: Learn AWS for Free Without a Credit Card

AWS is removing a major barrier to cloud onboarding by offering completely free, no-credit-card sandbox environments with automatic cleanup. This allows beginners to practice hands-on with AWS services without risk.

Patreon Blocks Crawlers From Stealing Creators' Work for AI Training

Patreon and Cloudflare have partnered to deploy live network-level blocking of AI training crawlers across all creator posts. This gives creators direct control over consent for their work without requiring an opt-out.

AI policy, regulation & governance

China tightens DeepSeek and other AI models as US embraces cheaper Chinese LLMs

China is considering implementing stricter controls on the release of AI models to protect its proprietary technology. This comes as US companies increasingly adopt more affordable Chinese LLMs like DeepSeek.

AI Legislative Update: July 10, 2026 — Transparency Coalition. Legislation for Transparency in AI Now.

Illinois has enacted a landmark AI Safety Measures Act, and California is tracking over 30 AI bills. Several California bills are advancing through committees and scheduled for August hearings, covering various aspects of AI regulation.

Council of the EU gives AI Omnibus final green light

The Council of the European Union has given final approval to the Digital Omnibus amendments to the EU AI Act. Key changes include deadlines for synthetic content labelling (Article 50), high-risk AI systems (2027–2028), and new prohibitions on intimate imagery and CSAM generation.

EU AI Act Enforcement Is Here: Chatbot Rules Live, High-Risk AI Delay Now Binding Law

The EU AI Act enforcement machinery will activate on August 2, 2026, with chatbot disclosure and synthetic content marking becoming legally binding. Deadlines for high-risk AI conformity assessments have been extended to December 2027 and August 2028.

The government’s secretive AI threat to creatives

Allegations suggest the Australian government is secretly considering weakening Australian copyright protections for AI training. This is reportedly in exchange for $220 billion in data centre investment.

Policy approaches to AI getting tweaked to allay public concerns

The Australian government is reportedly shifting towards a more interventionist approach to AI regulation. This includes demanding local community and employment benefits from tech giants before approving new data centers.

Google's New AI Training Setting Puts Australia on Notice

Google's opt-out-by-default data collection for AI training via user-uploaded media through Search, Lens, and Translate is creating corporate data governance risk in Australia. This precedes new Privacy Act disclosure obligations in December 2026.

US news media organizations have filed a federal sanctions motion against OpenAI. They allege concealment and destruction of evidence regarding ChatGPT's training on copyrighted news content, with one plaintiff's litigation costs exceeding $28 million.

Industry & market moves

Anthropic names former Fed Chair Bernanke to its independent trust

Anthropic has appointed former Federal Reserve Chair Ben Bernanke to its independent governance trust. He will advise on the economic impact of AI.

Google DeepMind director behind Gemini leaves for Hong Kong university after 20 years in US

Cao Liangliang, a director at Google DeepMind and contributor to the Gemini project, is returning to Hong Kong to lead AI research at Hong Kong Polytechnic University after two decades in the US.

Mercor is in talks for a $20B valuation

AI training startup Mercor is reportedly in talks for a $20 billion valuation funding round. This would be double its October Series C valuation of $10 billion, following a reported $2 billion annualized revenue run-rate and the acquisition of Deeptune.

Rackspace & Palantir launch sovereign AI framework

Rackspace and Palantir have jointly launched a managed framework specifically targeting regulated industries such as healthcare, financial services, and energy. This framework is designed for sovereign AI deployment with stricter data governance and operational oversight.

Venice AI Closes $65M Series A at $1B Valuation, Betting on Privacy-Focused, Uncensored AI Access

Venice AI, a privacy-focused AI platform, raised a $65 million Series A funding round at a $1 billion valuation. The company routes queries through over 200 models with optional encryption and claims 3 million active users and $70 million in annualized run-rate revenue.

Thrive Holdings Raises $2 Billion to Automate the Firms It Buys

Thrive Holdings has raised $2 billion in Series funding to acquire and automate service firms, rather than investing directly in AI startups. This strategy marks a significant bet on white-collar job automation in sectors like accounting and IT.

RecordPoint launches global partner program as AI adoption drives surge in data governance demand

RecordPoint is shifting to a channel-first business model with the launch of a structured four-tier reseller program and a referral track. This move aims to capitalize on surging enterprise demand for data and AI governance solutions.

AI product & feature launches

Alpha Bank expands AI partnership with ElevenLabs for voice banking

Alpha Bank has become the first bank in Greece to deploy ElevenLabs' voice technology on a large scale for customer service. This expansion integrates ElevenLabs' ElevenAgents voice assistant.

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

Cognition's SWE-1.7 demonstrates that applying reinforcement learning (RL) on top of an already extensively RL-trained base model can yield significant gains in code generation. This challenges the assumption of a post-training ceiling.

WitnessAI Unveils NER-D, a Breakthrough AI Model for Context-Aware Sensitive-Data Detection

WitnessAI has unveiled NER-D (Named Entity Recognition - Double pass), an AI model that achieves 20x faster detection speed than comparable generative methods. It also delivers benchmark-leading accuracy in context-aware sensitive data identification.

Research with immediate practical relevance

Tencent Open-Sources Hy3 Agent Model: How MoE Routing Cuts Costs, and What China Law Adds Back

Tencent has open-sourced Hy3, a 295B-parameter Mixture-of-Experts agent model with 21B active parameters, under Apache 2.0 licensing. Geographic carve-outs for EU, UK, and South Korea developers have been removed.