Agentic AI surges; frontier competition intensifies

Anthropic / Claude ecosystem

Anthropic announces 'programmatic credit pool' as agentic tool use rises

Anthropic introduces a new monthly‑refilling credit pool that separates agentic Claude usage from standard chat credits, reflecting growing agentic workloads.

Claude Code increases weekly limits by 50% through July 13, 2026

Anthropic temporarily boosts Claude Code usage limits by 50% in a promotional push against Codex, valid until July 13.

Frontier model providers

Building a safe, effective sandbox to enable Codex on Windows

OpenAI details the dedicated sandbox it built for Codex on Windows, providing safer isolation without relying on AppContainer or Windows Sandbox.

‘Gemini Spark’ is Google’s upcoming AI agent in the Gemini app

Google is preparing to launch a named AI agent called Gemini Spark within the Gemini app, featuring enhanced data sources and mixed permission behavior, beginning in beta.

DeepSeek unveils V4 AI with low costs, takes aim at OpenAI and Anthropic

DeepSeek officially launches its V4 model family, including V4 Pro, with a 1M‑token context window and claims of significant cost efficiency.

Google rolling out AI‑powered ‘Contextual suggestions’ on Android

Google introduces on‑device AI contextual suggestion cards for Pixel 10 devices on Android 16, offering proactive recommendations based on user context.

AI developer tooling & infrastructure

Introducing Semaphore for AI Agents: an AI‑native CI/CD experience

Semaphore launches an AI‑native CI/CD interface that lets AI agents debug and automate pipelines directly from coding environments.

GitHub Copilot introduces new Max $100 plan and Flex credit model

GitHub launches a new Copilot Max plan at $100/month with $200 total credits and a dual‑credit system to better serve heavy AI‑assisted development users.

Cloud & platform providers

AWS Enables AI Agents to Drive WorkSpaces Desktops

Amazon WorkSpaces launches a public preview that allows AI agents to automate tasks on managed desktops via MCP and auditable screenshots.

AWS Expands AI Security with Full Code Scanning Preview

AWS Security Agent now offers an early preview of automated, full‑repository code scanning to improve DevSecOps workflows.

ServiceNow and Google Cloud introduce new enterprise AI agents

ServiceNow and Google Cloud launch AI agents that collaborate across platforms to autonomously detect, diagnose, and resolve issues to prevent outages.

Browser Run now on Cloudflare Containers, faster and more scalable

Cloudflare migrates its Browser Run service to Containers, increasing per‑minute concurrency and lifting concurrent run limits.

AI policy, regulation & governance

AI preemption debate in Congress’s hands

A bipartisan push for federal preemption of state AI laws gains momentum, aiming to create a single nationwide regulatory framework.

Concerns raised over proposed Irish AI law’s independence and transparency

The Irish Human Rights and Equality Commission warns that the Regulation of Artificial Intelligence Bill 2026 may lack independent oversight and sufficient human‑rights protections.

AI education guidelines updated by European Commission

The European Commission revises its AI education guidelines to cover generative AI, student data, high‑risk systems, transparency, and human oversight.

UK Sovereign AI fund supports Isomorphic Labs

The UK government’s Sovereign AI fund invests in Alphabet’s Isomorphic Labs to advance AI‑driven drug discovery.

Industry & market moves

Anthropic forms $200 million partnership with the Gates Foundation

Anthropic announces a $200M partnership with the Bill & Melinda Gates Foundation, with the largest portion funding AI‑enabled health intelligence and public goods in low‑ and middle‑income countries.

Anthropic turns to SpaceX for Claude capacity

Anthropic partners with SpaceX to access more than 300 MW and 220k GPUs via the Colossus 1 facility, expanding Claude compute capacity to handle higher usage and reduce rate limits.

PwC and Anthropic expand alliance for enterprise agentic AI

PwC and Anthropic expand their alliance to deploy Claude‑based tools at scale and train tens of thousands of PwC professionals.

Microsoft eyes AI startup acquisitions to reduce reliance on OpenAI

Microsoft is reportedly considering buying AI startups such as Cursor and Inception to diversify its AI capabilities and lessen its dependence on OpenAI.

Parallel Web Systems raises $100M at $2B valuation for AI agent infrastructure

Parallel Web Systems, an AI agent infrastructure startup, closes a $100M Series B led by Sequoia at a $2B valuation.

Accelerating Australia and New Zealand's AI ambitions built on 20 years of AWS innovation

AWS highlights an $18.9B cumulative GDP‑investment in ANZ cloud/AI since 2012 and announces a new $20B expansion to build national AI capability.

Akamai to acquire LayerX for $205M to advance AI workforce security

Akamai Technologies announces a $205M acquisition of LayerX to integrate AI usage controls into its Zero Trust security portfolio.

Boomi announces new partnerships and acquisition to expand agentic AI capabilities

Integration platform Boomi unveils multiple partnerships and a planned acquisition (Lunar.dev) to enhance data activation and agentic AI connectivity.

Recursive raises $650M at $4.65B valuation for self‑improving AI research

Recursive, a self‑improving AI research company, secures $650M from GV, Greycroft, AMD Ventures, and NVIDIA at a $4.65B valuation.

OpenAI is reportedly exploring legal action over Apple’s alleged non‑fulfillment of a Siri integration deal, including possible breach‑of‑contract notices.

Department of War selects Applied Intuition to build enterprise autonomy pipeline

The U.S. Department of War awards Applied Intuition a contract to create a shared, production‑grade autonomous systems pipeline called Autonomy Factory.

AI product & feature launches

Tencent Cloud to end calls for 3 DeepSeek models on May 22

Tencent Cloud deprecates DeepSeek‑V3‑0324, V3.1‑Terminus, and R1‑0528, offering replacement versions with preserved output quality and latency.

Microsoft turns Copilot Studio into an AI agent control center

Microsoft expands Copilot Studio with governance roles and generally releases Agent 365, creating a centralized hub for monitoring and securing AI agents.

Microsoft retiring Copilot Mode on Edge as Copilot features become permanent

Microsoft removes the separate Copilot Mode in Edge, permanently integrating all Copilot functionalities into the browser across desktop and mobile.

SAP introduces Joule Work for the autonomous enterprise

SAP unveils Joule Work, a dynamic workspace that uses AI agents to automate end‑to‑end business processes across systems in real time.

Carta acquires AI‑based legal platform Avantia to launch Carta Law, an integrated legal/compliance system for private capital workflows.

AI robot performs 8‑hour shift better than humans in wild live stream

Figure AI’s Helix‑02 humanoid robot completes a full work shift autonomously and coordinates with other robots without human intervention.

8x8 AI Studio adds OpenAI GPT Realtime 2 for production voice agents

8x8 integrates GPT Realtime 2 into its AI Studio, giving contact centers a model with GPT‑5‑class reasoning, 128K context, and improved tool‑calling.

IBM introduces forward deployed units to make AI work in the real world

IBM Consulting launches a new delivery model that embeds senior teams of human experts and AI agents to move enterprise AI projects from pilot to production.

Research with immediate practical relevance

CAISI evaluation: DeepSeek V4 Pro nears GPT‑5‑level performance, but still behind

The U.S. AI Standards and Innovation Center (CAISI) rates DeepSeek V4 Pro as the best Chinese AI model, yet roughly 8 months behind the latest U.S. models like GPT‑5.

Coding Agent Index: Cursor tops first cross‑stack benchmark

Artificial Analysis releases its first coding agent benchmark ranking model‑and‑harness combinations; Cursor (Claude Opus 4.7) leads, highlighting the importance of the agent harness.

Humans are bad at making complex decisions; AI can call them out

Cornell researchers develop an Interactive Explainable Ranking tool that helps users uncover inconsistencies in their decision criteria, improving fairness without full AI autonomy.

AI models are getting better at replacing cybersecurity pros on certain tasks

The UK AI Security Institute finds that frontier models like Mythos Preview and GPT‑5.5 are quickly shortening the time needed to complete certain cybersecurity tasks.