AI policy surge; agents reshape enterprise

Anthropic / Claude ecosystem

No significant new developments.

Frontier model providers

OpenAI Introduces 'Dreaming' for Better ChatGPT Memory

OpenAI announced a new memory architecture called 'Dreaming' that enhances ChatGPT's ability to recall context over long interactions, improving personalization and task continuity with better compute efficiency.

OpenAI Adds Active Session Visibility for ChatGPT Security

OpenAI rolled out a new feature allowing administrators to view and audit active sessions for ChatGPT, improving governance and risk management for enterprise deployments.

Meta Business Agent Launched Globally Across Messaging Platforms

Meta unveiled the Meta Business Agent, an AI-powered assistant available on WhatsApp, Instagram, and Messenger, along with a platform for businesses to build and deploy custom AI agents for customer interactions.

Meta Delays Launch of Muse Spark AI Model for Developers

Meta has delayed the API launch of its multimodal Muse Spark model, citing ongoing testing with early partners and no confirmed release date.

GPT-5.5 Claimed to Disprove Decades-Old Sum-Product Conjecture

OpenAI's Sebastien Bubeck said that GPT-5.5, using a specific prompting workflow, disproved the sum-product conjecture, a long-standing mathematical problem, though independent verification is pending.

Google launched the AI Edge Gallery for macOS, allowing users to run Gemma 4 12B and other models locally on Macs, along with an on-device dictation app called Eloquent.

AI developer tooling & infrastructure

GitHub Launches Copilot Desktop App for AI Agent Coordination

GitHub released a Copilot Desktop app designed as an agent-native workspace to coordinate multiple AI coding agents across repositories, centralizing development lifecycle management.

Cloud & platform providers

xAI Grok Models Fully Integrated into Cloudflare AI Gateway

xAI and Cloudflare deepened their partnership, making the full Grok model suite available on Cloudflare’s AI Gateway with unified billing and management for developers.

AI policy, regulation & governance

Bipartisan US House Bill Seeks Federal AI Preemption

Representatives Obernolte and Trahan released a draft bill, the Great American AI Act, proposing a federal framework to regulate frontier AI models and preempt some state laws, with $300M for a centralised AI safety institute.

India Supreme Court Proposes Ban on AI in Judicial Verdicts

A draft regulation from India's Supreme Court committee would prohibit AI from making verdicts or bail decisions, ensuring human judges retain final authority and keeping AI purely advisory.

Sam Altman Urges Congress to Fund AI Testing, Not Mandate Approvals

OpenAI CEO Sam Altman testified before Congress, recommending increased government funding for AI testing rather than requiring model pre-approval, preserving voluntary evaluation frameworks.

OpenAI, Anthropic Sign Letter Urging US Action on AI Bioweapons

Leading AI companies including OpenAI and Anthropic signed a letter calling for new US laws to require DNA synthesis screening, aiming to prevent AI-enabled biological threats.

Industry & market moves

Anthropic and University of Tokyo Partner to Measure AI Impact

Anthropic and the University of Tokyo are forming a partnership to create an index measuring how generative AI affects business, government, and education in Japan.

Cohere Denies Revenue-Generating Partnership with Palantir

Cohere publicly denied having a commercial relationship with Palantir, clarifying that any pilot collaboration was short-term and did not generate revenue, contradicting recent reports.

xAI Pauses Specialist Hiring for Grok Training

xAI has paused the hiring of domain specialists to train the Grok chatbot, signaling a possible shift in its training strategy or resource allocation.

Lovable Expands Google Cloud Collaboration for AI-Powered Software Creation

Lovable expanded its collaboration with Google Cloud to scale AI-powered software creation for enterprises, aiming to serve millions of users globally with accessible development tools.

Cloudflare Acquires VoidZero to Build AI-Native Web

Cloudflare acquired VoidZero, the team behind Vite, to integrate its tooling into Cloudflare's edge and Workers platform, and committed a $1M ecosystem fund for open-source maintainers.

Alnylam Signs $2B AI Drug Design Deal with Inceptive

Alnylam Pharmaceuticals entered a collaboration worth up to $2 billion with Inceptive to use AI for designing siRNA therapies, paying $30 million upfront.

Nvidia Acquires Predictive AI Startup Kumo AI

Nvidia acquired Kumo AI, a startup known for accurate predictive models, to bolster its AI software portfolio for enterprise customers.

CNTXT AI Acquires Actualize to Expand Arabic Voice Agents

UAE-based CNTXT AI acquired Actualize to strengthen its sovereign Arabic AI stack with GCC-dialect-native voice models and engineering talent.

Palo Alto Networks Completes Portkey Acquisition to Secure AI Agents

Palo Alto Networks closed its acquisition of Portkey, integrating its AI gateway into the Prisma AIRS platform to enhance security for enterprise AI agents at scale.

CXAI Acquires EngineRoom to Expand Revenue Analytics

CXApp Inc. acquired EngineRoom to enhance its AI-driven revenue growth analytics platform, boosting annualized revenue and EBITDA.

KPMG Partners with Anthropic to Deploy Claude to 276,000 Staff

KPMG embedded Claude across its Digital Gateway platform, providing all 276,000 employees with AI assistance and planning to extend it to client services, focusing on trust and private equity.

AI product & feature launches

Trend Micro to Use Anthropic's Claude Mythos for Cyber Defense

Trend Micro announced plans to integrate Anthropic's Claude Mythos into its corporate cyber defense services for rapid vulnerability detection and remediation, marking the first Japanese company to publicly disclose access to Mythos.

Gopuff Launches AI Shopping Assistant Go Powered by SpaceXAI

Gopuff introduced Go, an AI shopping assistant that uses SpaceXAI's frontier reasoning to deliver personalized recommendations and faster order processing through delivery intelligence.

Harness May 2026 Update Adds Claude MCP Server Integration

Harness released over 60 new features, including an MCP server that integrates with Claude, enabling AI-native development with real-time pipeline access and safety guardrails.

MWM AI Partners with Google Cloud to Launch AI Mobile Squad

MWM AI announced a collaboration with Google Cloud to launch AI Mobile Squad, a platform leveraging Gemini Enterprise to help app creators and entrepreneurs build AI-powered mobile agents.

Cognizant Deploys Snowflake Cortex AI Agents for Enterprise Efficiency

Cognizant announced it is building intelligent agents on Snowflake's Cortex platform, powered by Snowflake CoCo, to accelerate enterprise AI adoption and improve operational efficiency.

Xense Robotics Showcases Full-Stack Tactile Intelligence at ICRA

Xense Robotics demonstrated a complete tactile intelligence system at ICRA 2026, including perception, world modeling, and vision-touch-language-action models for robots.

NVIDIA NemoClaw Blueprint for Secure Autonomous Simulation Engineers

NVIDIA showcased NemoClaw, an open blueprint for deploying secure, autonomous AI agents that can accelerate industrial simulation workflows from weeks to hours.

Amazon Develops Proteus Warehouse Robot with Language Interface

Amazon is testing a next-generation Proteus robot that workers can speak to via natural language, with plans for European pilot deployments by 2027.

Asana Launches AI-Powered Suite for Human-Agent Work Management

Asana introduced Agentic Work Management, including a chief-of-AI staff, three new applications, and AI teammates, aiming to unify task coordination between humans and AI agents.

Research with immediate practical relevance

Google Releases Magenta RealTime 2: Open Music AI with On-Device Inference

Google's Magenta project launched RealTime 2, an open-weights live music model enabling real-time musical interactions with 40ms latency, running locally on devices.

Free AI Model Powers Self-Spreading Worm in Enterprise Test

Security researchers demonstrated that an open-weight AI model from 2025 could be used to create a self-spreading worm in enterprise test environments, exploiting known vulnerabilities at scale.