Agents drive infra, regulation; Anthropic builds chips

Anthropic / Claude ecosystem

Anthropic Confirms In-House Chip Team to Power Claude

Anthropic publicly confirmed for the first time that it is building an in-house silicon team to co-design hardware and models for Claude, signalling a shift toward custom chips while maintaining reliance on third-party hardware.

Anthropic built an inspection layer that lets enterprises block sensitive data before it reaches Claude

Anthropic extends DLP enforcement to all Claude Enterprise surfaces (chat, Code, Cowork, MCP connectors) via webhook-based inference hooks, letting enterprises block sensitive data before it reaches the model.

Frontier model providers

New ways to learn and teach with ChatGPT Work and Codex

OpenAI introduces three role-specific education plugins for ChatGPT that help students and educators leverage agentic AI capabilities with integrated access to institutional tools and course materials.

Cohere bets on enterprise-optimised sovereign AI, targets 1 trillion-parameter model this year

Cohere targets a 1 trillion-parameter sovereign AI model optimized for enterprise deployment across cloud and on-premise environments, emphasizing computational efficiency and vendor-agnostic platform strategy.

AI developer tooling & infrastructure

VS Code 1.132 Sharpens Agent Workflows with Side Chats, Browser Comments

VS Code 1.132 introduces side chats for agent workflows and element-level browser comments, enabling developers to ask contextual questions without interrupting active agent turns.

1-Click RCE Flaw in Cursor, VS Code, and Google Antigravity Exposes 50M Developers to Cyberattacks

Critical one-click RCE flaw affecting 50 million developers across three major code editors has been patched, but highlights rapid propagation of security flaws in AI-native developer tools.

AWS’s Kiro Crew aims to turn AI coding agents into autonomous engineering teams

AWS released Kiro Crew, an open-source orchestration platform that coordinates multiple AI agents to enable autonomous, long-running engineering workflows across repositories and developer tools with persistent memory and human-approval governance.

Scaling AI Agent Infrastructure with the MCP Stateless updates

Google led the charge to decouple MCP from stateful transport constraints, enabling stateless protocol core that scales on ordinary HTTP load-balanced infrastructure without session pinning or Redis stores.

Uber open-sources ADR, an agent-native security framework for observability, benchmarking, and threat detection, signaling that AI safety has shifted from research to production operations.

Query Dynatrace Managed data directly in MCP-compatible clients

Dynatrace Managed MCP Server now enables AI assistants (Claude, GitHub Copilot, Kiro) to query self-hosted Dynatrace observability data in natural language, closing the gap between SaaS and Managed deployment models.

RouterBase Unified AI API Now Connects 200+ Frontier

RouterBase now aggregates 200+ frontier AI models including text, image, video and audio through a single OpenAI-compatible API with automatic failover and unified billing.

Microsoft introduces open-source framework for AI agents

Microsoft releases Orchard, an open-source framework designed for building scalable agentic AI systems.

Runware launches portable AI data center; Cloudflare simplifies debugging

Modular data center for portable AI infrastructure testing and simplified debugging for cloud-native AI workloads.

Cloud & platform providers

WriteGuard: Fine-grained controls for MCP Servers

Cloudflare launches WriteGuard, a policy and auditing layer for MCP Servers that enables fine-grained control over agent write actions with centralized attribution and audit trails.

Introducing Web Search on Amazon Bedrock for foundation model grounding

Web Search is now natively integrated into Amazon Bedrock as a server-side tool, enabling foundation models to ground responses in current web knowledge without third-party vendor integration.

Mainframe migration and modernization with AI

Google Cloud combines Gemini AI models with mainframe-specific modernization products spanning assessment, modernization, de-risking, and data migration to enable safe, iterative mainframe-to-cloud migration at enterprise scale.

Advance Zero Trust for AI: New tools and guidance to secure AI agents and DevSecOps

Microsoft releases new Zero Trust for AI guidance, assessment tools, and DevSecOps workshop pillar to help organizations secure AI agents and agentic systems at scale.

Cloudflare OS: an open platform for agents, apps, and work

Cloudflare open-sourced a new version of Cloudflare OS, an agent workspace platform with built-in security governance and app-building capabilities for enterprise organizations.

AI policy, regulation & governance

DeepMind and Isomorphic Labs Unveil Bioresilience Framework

DeepMind and Isomorphic Labs launched a bioresilience framework gating access to frontier biological AI models via four-tier pipeline, DNA watermarking, and restricted deployment to vetted institutional partners to manage CBRN risks.

Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

Meta's ad review systems failed to catch over 50 ads containing AI-generated child sexual abuse material that were live across Facebook, Instagram, Messenger, and Threads for months.

AI Leaders Propose SAFE Guidelines for Cybersecurity Transparency

Open Secure AI Alliance members propose SAFE guidelines to enable confidential collection and analysis of agentic AI cybersecurity incidents for ecosystem-wide threat intelligence sharing.

Guidelines on transparency obligations for providers and deployers of certain AI systems

EU Commission publishes binding transparency guidelines under Article 50 of the AI Act, requiring AI providers to disclose AI interactions and deployers to inform users of emotion recognition, deepfakes, and AI-generated text.

Britain Says it Is Open to AI Regulation if Voluntary Safeguards Fall Short

UK signals willingness to introduce AI regulation if voluntary pre-deployment testing by its AI Security Institute proves insufficient to manage frontier model risks.

White House AI Framework Excludes Open-Weight Models From Federal Security Review, Creating Structural Competitive Asymmetry

The federal framework excludes open-weight models from security review, creating a regulatory asymmetry that favors rapid open-source iteration over closed-source vendors subject to 30-day pre-release evaluation.

TGA flags compliance action on AI scribes as review finds safety concerns

Australian regulator shifts from education to enforcement against AI scribe vendors for scope creep, lack of transparency, and inadequate post-market monitoring of medical devices.

Australian PM signals intent to mandate licensing (not scraping) for AI training on copyrighted creative works, prioritising artist control over price and usage.

Australian government clarifies which AI development activities—model training, testing, and integration—qualify for R&D tax offsets, excluding routine implementations and parameter tuning.

Industry & market moves

Anthropic signs $10 billion deal with AI cloud startup Volta

Anthropic secures long-term compute capacity from Volta via a $10B six-year deal, powered by Nvidia Vera Rubin chips in a Norwegian data center as part of its ongoing effort to expand compute infrastructure.

The next chapter of our AI momentum

Google reorganizes DeepMind leadership with Demis Hassabis focusing on AGI strategy as Chair and Chief Scientist while Koray Kavukcuoglu assumes operational control of Google DeepMind; veteran researcher Jeff Dean departs to launch independent AI research venture.

Cohere establishes Korea unit to lead APAC push with sovereign AI

Cohere establishes Korea as its APAC hub and plans to double the region's workforce by year-end, betting on demand for sovereign AI in regulated markets.

DeepSeek reportedly reopens talks on a RMB50 billion second funding round

DeepSeek reopens fundraising talks for a second round at RMB50 billion valuation, potentially reaching agreement by late August.

Bending Spoons has entered into a definitive agreement to acquire Airtable for $1.285 billion

Bending Spoons acquires Airtable for $1.285 billion in an all-cash deal, marking its first acquisition since going public on Nasdaq in July 2026.

United States - AlixPartners Acquires Leading Agentic AI Consulting Firm, Artium

AlixPartners acquires agentic AI consulting firm Artium to combine AI engineering capabilities with 45 years of industry consulting expertise and expand enterprise AI agent deployment offerings.

Google Is in Talks for a $1.5 Billion-Plus Deal With Mechanize

Google is in early-stage talks to acquire San Francisco AI coding startup Mechanize for over $1.5 billion, primarily to absorb its engineering talent for improving Google's AI coding capabilities.

BigHand's acquisition of Ayora combines AI-powered legal pricing expertise to accelerate AI innovation and market position in legal technology.

Lightspeed-backed Multiplier raises $35m to expand professional services platform

Multiplier, a Lightspeed-backed professional services platform, closes $35M Series B to accelerate acquisitions and develop AI tools for accounting firms.

QuantHealth Snags $45M to Simulate Clinical Trials Before Patients Ever Enroll

Israeli clinical trial simulation startup QuantHealth raises $45M Series B to expand AI-driven drug development platform that predicts patient outcomes with up to 90% accuracy.

VC firm Sorin Investments leads ₹30 crore funding round in AI firm Kily

Sorin Investments-led ₹30 crore round in agentic AI startup Kily demonstrates investor confidence in autonomous commerce execution for Indian e-commerce brands.

Nvidia to build Starmind AI1 satellite compute payload for SpaceX

SpaceX and Nvidia jointly announced development of the Starmind AI1 satellite to perform advanced AI computing in low Earth orbit, with plans for a constellation of up to one million satellites.

AI product & feature launches

Meta makes AI Developer Assistant its primary support option

Meta makes its AI Developer Assistant the primary support option for developers building on its platform, replacing traditional support tickets and documentation navigation.

Tenable Advances Exposure Management with Coverage Across Every Major AI Platform and Developer Tool

Tenable expands its AI exposure management platform to cover Google Gemini, Model Context Protocol deployments, and AI-native IDE tools, unifying visibility across all major LLMs and development tools.

Seamless.AI Partners With ChatGPT To Unite Verified Sales Data and Prospecting Right Inside the Chatbot

Seamless.AI's native ChatGPT integration brings real-time B2B sales data and prospecting automation directly into OpenAI's assistant, enabling users to move from a prompt to verified leads in seconds without leaving the chatbot.

Skunk Works® Advances Sensor-Powered AI Fighter Intercept

Demonstrated closed-loop AI combat autonomy using real sensor data on operational fighter aircraft, achieving 27 successful intercepts in eight flights.

Hark previews its browser use agent for completing tasks

Hark Handoff is a browser-use agent that navigates websites without official APIs and claims to be faster and cheaper than competing models like GPT-5.5 and Opus 4.8.

Research with immediate practical relevance

XunZi, an AI biologist, reveals disease-modifying targets

XunZi establishes a paradigm-shifting framework to translate fragmented biomedical knowledge and data into actionable therapeutics, identifying disease-modifying targets such as CHK2 and IRAK4 in Parkinson's disease with experimental validation.