The Informed Layer · Newsscan
Daily briefing

Frontier labs push agents; global AI regulation surges

← All issues

Anthropic / Claude ecosystem

Anthropic expands Claude enterprise security with self-hosted agents and KPMG partnership

Anthropic has introduced new enterprise security features for Claude, including self-hosted deployment options, MCP tunnels, and Cloudflare sandbox integration, along with a partnership with KPMG to scale enterprise AI offerings.

Claude Managed Agents adds Dreaming, Multiagent Orchestration, and Outcomes features

Anthropic has launched three new features for Claude Managed Agents: Dreaming for goal exploration, Multiagent Orchestration for coordinating multiple agents, and Outcomes for production-grade performance tracking, now in public beta.

Frontier model providers

OpenAI launches Guaranteed Capacity program for enterprise compute

OpenAI is offering a new enterprise program that guarantees compute capacity across cloud providers with annual spend commitments, addressing capacity constraints for large-scale AI adoption.

Google unveils Gemini for Science, an AI research toolkit and platform

Google has launched Gemini for Science, an integrated AI-powered platform combining multiple DeepMind technologies to accelerate scientific discovery, featuring science skills unified across 30+ life science data sources.

Meta introduces AI Incognito Chat with promise of unreadable conversations

Meta has launched AI Incognito Chat, a feature that processes AI queries in a secure environment ensuring conversations cannot be read by Meta or WhatsApp, enhancing privacy.

Meta launches Muse Spark, its most sophisticated AI model

Meta has unveiled Muse Spark, described as its most advanced AI model to date, likely a new iteration in its Llama family, though details are limited.

Meta unveils custom AI chip to rival Nvidia H100

Meta has announced a new internally developed AI accelerator designed for large-scale recommendation and Llama-family models, aiming to compete with Nvidia's H100 on performance and cost.

Cohere open-sources Command A+ with lossless quantization and native citations

Cohere has released Command A+, its strongest model, under Apache 2.0 license, featuring lossless quantization and native citation generation for grounded, traceable facts in enterprise applications.

Google launches Antigravity 2.0, a standalone multi-agent development suite

At Google I/O, Antigravity 2.0 was introduced as a standalone desktop app for orchestrating multi-agent workflows with scheduling and voice features, along with a public SDK and Gemini API integration, marking a major push into agentic AI tooling.

Google unveils new AI-powered ad agents with Gemini backing

Google has launched a new suite of AI-powered advertising tools, including Gemini-backed agents and explainers, aiming to automate and enhance digital marketing campaigns.

AI developer tooling & infrastructure

VS Code 1.121 adds remote agent sessions and HTML/Mermaid previews

Visual Studio Code version 1.121 introduces remote agent sessions and built-in HTML and Mermaid previews, enhancing AI-assisted and agent-oriented development, particularly with Claude Code integrations.

Chrome introduces WebMCP framework to expose website tools to AI agents

Chrome is rolling out WebMCP, a browser-based framework that allows websites to expose tools to AI agents via declarative or imperative APIs, reducing brittleness in agent-website interactions, available as an origin trial in Chrome 149.

1Password integrates with OpenAI Codex for just-in-time credential injection

1Password has introduced an Environments MCP Server for OpenAI Codex, enabling AI coding agents to access secrets without exposing them in prompts or code, enhancing security.

SISTRIX opens MCP server access to all subscription tiers via OAuth

SISTRIX has made its MCP server available to all customers without requiring API keys or credits, using OAuth, expanding AI tool access for SEO data.

Resolve AI introduces multi-agent investigation system for production root cause analysis

Resolve AI has expanded its platform with a multi-agent investigation system and shared real-time collaboration to improve root cause analysis in production environments stressed by AI-generated code.

Cloud & platform providers

Google Cloud launches Agent Substrate for ultra-scale agent workloads on GKE

Google Cloud introduced Agent Substrate to improve real-time scheduling and data locality for large-scale AI agent workloads, built on GKE with Agent Sandbox.

Microsoft releases Azure Linux 4.0, optimized for AI workloads

Microsoft has released Azure Linux 4.0 and Azure Container Linux, providing AI-optimized, secure operating system layers for cloud workloads, alongside an open agentic AI ecosystem.

AI policy, regulation & governance

Taiwan establishes National AI Strategy Committee and Artificial Intelligence Basic Act

Taiwan's Executive Yuan has announced the formation of a National Artificial Intelligence Strategy Special Committee and plans to implement an Artificial Intelligence Basic Act with risk assessments and governance rules.

Singapore proposes AI nutrition labels for consumer AI products

Singapore is engaging tech firms to develop a voluntary global-first 'nutrition label' system for AI products, indicating appropriate and inappropriate uses, starting with consumer applications.

Singapore updates AI governance framework for agentic AI

Singapore's IMDA has released an updated Model AI Governance Framework (MGF) specifically for agentic AI, incorporating real-world case studies from over 60 organizations and new best practices for responsible deployment.

EU Commission publishes draft guidance on high-risk AI system classification

The European Commission released draft guidelines to help determine which AI systems fall under the high-risk category of the EU AI Act, with practical sector examples and public feedback invited.

Trump to sign executive order on voluntary AI oversight for frontier models

The White House is expected to sign an executive order establishing a voluntary 90-day disclosure framework for frontier AI model developers, amid security concerns from conservative lawmakers.

Bipartisan revised NO FAKES Act targets unauthorized AI digital replicas

A bipartisan group in the US Senate has reintroduced the NO FAKES Act, aimed at prohibiting non-consensual digital replicas of voice and likeness and preempting state laws, addressing deepfake concerns.

China issues first national framework for agentic AI

China's cyberspace administration has published Implementation Opinions on the Standardised Application and Innovative Development of Intelligent Agents, the world's first comprehensive statutory framework for agentic AI, covering governance, safety, and industrial strategy.

Australia's APRA warns banks on AI governance gaps

The Australian Prudential Regulation Authority (APRA) has issued supervisory findings and an industry letter warning that governance and assurance practices in financial services are not keeping pace with rapid AI adoption, signaling increased scrutiny.

Australia to appoint Chief AI Officers across public service

Australia is establishing Chief AI Officer roles across its public service to close governance gaps in AI adoption and improve accountability, following budget commitments.

Industry & market moves

Anthropic to pay SpaceX $45 billion over three years for AI compute

Anthropic has agreed to pay SpaceX approximately $45 billion over three years for computing resources to support the development of its Claude AI models, marking one of the largest compute deals.

Anthropic and KPMG sign alliance to deploy Claude to 276,000 staff

KPMG and Anthropic have formed a global alliance to integrate Claude across KPMG's 276,000 employees, launching a Digital Gateway powered by Claude for tax and private equity services.

OpenAI plans confidential IPO filing with Goldman Sachs and Morgan Stanley

OpenAI is preparing to file confidentially for an IPO as early as Friday, aiming for a public debut possibly in September, which could be the largest public offering in recent memory.

Google and Singapore deepen AI partnership for public good

Google has extended its partnership with the Singapore government to advance the National AI Strategy, focusing on AI impact, workforce readiness, and public-good applications.

Meta turns internal operations into AI post-training lab with Scale AI

Meta is converting its internal processes into a large-scale post-training lab, blending internal feedback with external evaluation from Scale AI to refine its AI models.

DeepSeek forms Harness team to build code agent rivaling Claude Code

DeepSeek is assembling a dedicated Harness team to develop DeepSeek Code, a commercial coding agent tool aimed at competing with Anthropic's Claude Code and OpenAI's Codex, potentially lowering costs with its low-cost inference.

SpaceX plans to acquire AI coding startup Cursor for $60B post-IPO

SpaceX intends to acquire Cursor, an AI coding assistant startup, for $60 billion in a deal closing within 30 days of SpaceX's IPO, with a $10B breakup fee, subject to regulatory approval.

Anthropic explores Microsoft Maia AI chips for Claude inference

Anthropic is evaluating running Claude models on Microsoft's Maia 200 AI accelerator, which promises a 30% inference cost reduction, potentially reducing reliance on Nvidia.

AI-native enterprise services firm backed by Anthropic and Blackstone acquires Fractional AI

A newly formed AI-native enterprise services company, backed by Anthropic, Blackstone, and Hellman & Friedman, has acquired Fractional AI to expand its AI service offerings.

Hark raises $700M Series A at $6B valuation for universal AI interface

Stealth startup Hark raised a massive $700 million Series A at a $6 billion valuation to develop a secretive universal AI interface, drawing significant investor interest.

Kawasaki Heavy and Nvidia partner on physical AI, opening US robot center

Kawasaki Heavy Industries and Nvidia will form a joint development center in Silicon Valley to fuse robotics with physical AI, accelerating industrial automation.

Innovaccer acquires CaduceusHealth to expand AI-powered healthcare platform

Healthtech unicorn Innovaccer has acquired CaduceusHealth, a US-based revenue cycle management firm, to integrate AI-driven analytics into its Flow platform and expand healthcare service offerings.

Sonar acquires Gitar to add AI code review to its verification platform

Sonar has acquired Gitar, expanding its static analysis and code quality platform with AI-driven code review capabilities for better coverage and efficiency.

Grundium acquires Visiopharm for AI-powered precision pathology platform

Grundium has acquired Visiopharm to combine digital pathology imaging with AI-driven diagnostics software, creating an end-to-end platform for clinical and research use.

Mythic acquires Videantis to boost hybrid AI compute with 100x energy efficiency goal

Mythic has acquired Videantis to integrate its analog compute-in-memory technology with Videantis's digital IP, aiming for up to 100x energy efficiency improvement over conventional AI accelerators.

AI product & feature launches

Tenable partners with Anthropic for AI-driven exposure management

Tenable announced a partnership with Anthropic to integrate frontier AI capabilities into its Tenable One platform for improved exposure management.

DoiT launches AI-powered PerfectScale for autonomous cloud commitment optimization

DoiT has launched PerfectScale for Commitments, an AI-driven tool that autonomously optimizes cloud savings plans and committed use discounts across AWS, Google Cloud, and forthcoming Azure.

Kore.ai launches Artemis, an AI-native enterprise agent platform

Kore.ai introduced Artemis, a new AI-native platform for building, governing, and optimizing enterprise AI agents, initially on Microsoft Azure with plans for broader cloud availability.

Zendesk launches Autonomous Service Workforce with outcome-priced AI agents

Zendesk has introduced the Autonomous Service Workforce, a set of specialised AI agents that operate across channels and are priced based on resolutions, replacing traditional deflection bots.

Tenable launches Hexa AI for automated remediation across attack surfaces

Tenable has announced the general availability of Hexa AI, an agentic AI that automates exposure management and accelerates remediation actions across diverse attack surfaces.

QIAGEN launches QIA Agent, an AI assistant for scientific workflows

QIAGEN has launched QIA Agent, an AI-powered assistant that integrates product data, protocols, and support into a conversational interface across its sample-to-insight workflows.

AMD announces Ryzen AI Halo dev platform, taking on Nvidia Spark at $3999

AMD has introduced the Ryzen AI Halo developer platform, claiming competitive tokens-per-dollar performance for local AI inference up to 200B parameters, with pre-orders starting June 2026.

Research with immediate practical relevance

OpenAI model disproves 80-year-old Erdős conjecture in discrete geometry

An OpenAI reasoning model has disproven a central conjecture in discrete geometry by Paul Erdős, demonstrating that infinite constructions with superlinear unit-distance pairs exist, a major advancement in AI-assisted mathematics.