Agents multiply; Google previews RT-3; Regulators advance

Anthropic / Claude ecosystem

Anthropic Release Notes - May 2026 Latest Updates - Releasebot

Anthropic has released Claude Code 2.1.126 with OAuth improvements, project purge tools, and Windows/PowerShell fixes. The company is also retiring the 1M token context beta for Sonnet 4.5/4 in preparation for migration to 4.6.

Frontier model providers

Google DeepMind teases RT-3 with new capabilities - RobotWale News

Google DeepMind's RT-3 robotic foundation model shows significant advancements in generalization and dexterity for humanoid robots, with improved long-horizon planning and natural language understanding capabilities.

Google DeepMind Debuts Multilingual TTS Model Featuring Integrated SynthID Watermarking for Synthetic Voice Authentication | VoiceAI Pulse - Latest News & Insights in AI Voice Technology

Google DeepMind has integrated SynthID watermarking directly into its Gemini 3.1 Flash TTS model to authenticate AI-generated audio and prevent deepfakes.

Meta's Autodata: AI Agents Now Build Their Own Smarter Training Data - AI Universe: A News Startup

Meta's Autodata framework allows AI agents to autonomously generate and refine training datasets, significantly widening performance discrimination and potentially shifting AI advancement from compute power to data quality bottlenecks.

Google releases experimental ‘COSMO’ AI assistant on Play Store

Google accidentally released an experimental AI assistant named COSMO on the Play Store, featuring a local Gemini Nano model and multiple AI-powered productivity skills.

Google Unveils Powerful AI Agent, Boosts Cloud Capabilities – OVEX NEWS

Google released Gemini Deep Research, a new AI agent designed to accelerate scientific and financial research analysis from weeks to days by processing vast research literature and multimodal data.

AI developer tooling & infrastructure

GameMaker Adds Claude Code to GM-CLI for AI-Assisted Game Development

GameMaker has integrated Anthropic's Claude Code into its GM-CLI toolchain, making it the first mainstream game engine to natively ship an AI developer agent. This integration aims to assist with AI-driven game development alongside a source-available runtime.

JobRunr Introduces ClawRunr, an Open-Source Java AI Agent - InfoQ

JobRunr has introduced ClawRunr, an open-source Java AI agent designed for executing scheduled and recurring background tasks with built-in persistence and monitoring capabilities.

Cloud & platform providers

Amazon Bedrock AgentCore launches capabilities for optimizing agent performance in preview - AWS

Amazon Bedrock AgentCore now provides automated recommendations, batch evaluations, and A/B testing to validate and deploy AI agent performance improvements without manual intervention, completing the observe-evaluate-improve loop.

AI policy, regulation & governance

CAISI Evaluation of DeepSeek V4 Pro | NIST

NIST's Center for AI Standards and Innovation (CAISI) benchmark suite shows DeepSeek V4 Pro lags US frontier models by approximately 8 months on independent reasoning and agentic tasks, despite DeepSeek's claims of parity.

Colorado’s AI compromise would focus regulations on informing consumers when the technology is used  - The Colorado Sun

Colorado has introduced a compromise AI regulation bill (SB 189) that mandates consumer notification when AI makes consequential decisions, relaxes disclosure requirements on system mechanics, and delays implementation to January 2027.

US government, allies publish guidance on how to safely deploy AI agents | CyberScoop

US and Five Eyes cybersecurity agencies (CISA, NSA, and allies) have jointly published the first comprehensive guidance on securing autonomous AI agents, warning that the technology is already deployed in critical infrastructure with insufficient safeguards.

Senators take another swing at bill to codify federal AI resource | FedScoop

Bipartisan senators have reintroduced S.4441, the CREATE AI Act, to codify the National AI Research Resource (NAIRR) in statute. This aims to expand AI research access for universities, nonprofits, and students.

Clean up AI or go away: rules proposed for data centres - Yahoo News Australia

NSW planning authorities are being recommended to condition data center approvals on AI model legality guarantees and renewable energy commitments to enforce copyright compliance and environmental standards.

A Chinese appeals court ruled that a company cannot use AI replacement as justification for unlawful dismissal, signaling labor rights protection amid widespread AI adoption.

Oscars bans AI actors and writing from awards

The Academy of Motion Picture Arts and Sciences has formally banned AI-generated acting and writing from Oscar eligibility, explicitly stating that awards can only go to human-authored work.

Industry & market moves

OpenAI And Yubico Partner To Bring Custom Phishing-Resistant YubiKeys To OpenAI Users

OpenAI and Yubico have launched an industry-first program to distribute custom phishing-resistant hardware security keys, extending enterprise-grade YubiKey protection to ChatGPT users at scale.

New AI lab Core Automation 'nerdsniped' researchers from ...

Multiple top researchers from Anthropic, Google DeepMind, and OpenAI have joined Core Automation, a newly launched AI startup aiming to build the world's most automated AI lab.

Meta Acquires Robotics AI Startup As It Makes The Push Into Humanoid Machines

Meta has acquired robotics AI startup Assured Robot Intelligence (ARI) to advance humanoid robot control and self-learning capabilities for high-value labor markets.

Nebius agrees to acquire Eigen AI, strengthening Nebius Token ...

Nebius has acquired Eigen AI for $643 million to integrate inference optimization capabilities into its Nebius Token Factory, aiming to strengthen its position in production AI inference.

Encord Announces $60 Million in Series C Funding

Encord has successfully raised $60 million in Series C funding to scale its AI-native data infrastructure, driven by accelerating physical AI deployment.

AI product & feature launches

NAudio Modernization with Claude Code - Mark Heath

A developer demonstrated how Anthropic's Claude Code can be used to accelerate large-scale open-source modernization projects, including API redesign, performance optimization, and test coverage, without encountering token budget limitations.

OpenAI starts laying foundations for ChatGPT ads in EU

OpenAI is implementing consent-based tracking mechanisms to lay the technical groundwork for expanding its ChatGPT ads business into Europe, complying with stringent EU privacy regulations.

Microsoft spent two months overhauling Windows, and yes, that includes cutting Copilot

Microsoft is retreating from its aggressive Copilot integration strategy across Windows, removing AI buttons from core applications to restore user trust and make AI features more intentional.

Zensai Introduces Human Success Agent for Microsoft Agent 365, Redefining AI-Driven Learning and Human Success in the Flow of Work

Zensai has launched its Human Success Agent on Microsoft Agent 365, enabling enterprises to transform workforce data into real-time, actionable insights embedded directly within the Microsoft 365 workflow.

Citi launches Arc to scale AI agents across the business | CIO Dive

Citigroup has launched Arc, a new platform designed to deploy embedded AI agents at enterprise scale across all business lines and geographies, automating manual tasks such as research and data analysis.

Microsoft wants lawyers to trust its new AI agent in Word documents | The Verge

Microsoft has launched Legal Agent, a specialized AI tool for legal teams designed to review contracts and identify risks, leveraging expertise from acquired Robin AI engineers.

Kamiwaza AI Launches Kamiwaza 1.0: Secure AI Orchestration Built for Regulated Industries

Kamiwaza AI has launched Kamiwaza 1.0, providing governed AI collaboration and secure-by-design infrastructure for regulated industries without centralizing enterprise data.

Research with immediate practical relevance

A New NVIDIA Research Shows Speculative Decoding in NeMo RL Achieves 1.8× Rollout Generation Speedup at 8B and Projects 2.5× End-to-End Speedup at 235B - MarkTechPost

NVIDIA research demonstrates a 1.8× rollout generation speedup at 8B scale and projects a 2.5× end-to-end speedup at 235B scale through lossless speculative decoding integrated into the NeMo RL training loop.

Google’s Aletheia AI Solves 6 Research Math Problems | byteiota

Google DeepMind's Aletheia AI agent is the first AI to autonomously produce publishable mathematical research, solving 6 unseen research-level problems and generating a complete peer-reviewed paper without human intervention.