The Informed Layer · Newsscan
Daily briefing

Agents drive AI infrastructure & regulation; Funding surges

← All issues

Anthropic / Claude ecosystem

Introducing Claude Platform on AWS: Anthropic’s native platform, through your AWS account

AWS has become the first cloud provider to offer native access to Anthropic's Claude Platform, providing a unified billing experience, IAM authentication, and CloudTrail logging for enterprises. This integration allows AWS customers to directly access and manage Claude's AI services within their existing AWS accounts, streamlining operations and enhancing security.

Lyrie.ai Joins First Batch of Anthropic’s Cyber Verification Program - Cybersecurity Insiders

Lyrie.ai has been accepted into the inaugural batch of Anthropic's Cyber Verification Program, allowing for security research and red-team workflows on Claude infrastructure. Concurrently, Lyrie.ai released its Agent Trust Protocol (ATP) as an open cryptographic standard to enhance AI agent identity and authorization.

Frontier model providers

OpenAI Adds Codex Chrome Extension for Signed-In Web Tasks

OpenAI has released a new Codex Chrome Extension, enabling the AI to operate within signed-in browser sessions for services like Gmail, Salesforce, and internal dashboards. This is facilitated through supervised tab-group isolation and per-site approval gates, offering controlled interaction rather than unrestricted browser access.

AI developer tooling & infrastructure

Deepclaude keeps Claude Code experience, cuts model costs to one-seventeenth

Deepclaude, a new open-source tool, allows developers to utilize more cost-effective models like DeepSeek V4 Pro while maintaining the familiar coding-agent interface and workflow of Claude Code. This innovation significantly reduces operational costs for AI development, reportedly cutting them to one-seventeenth of Anthropic's pricing.

Meta Adds AI Agentic Workflow to Immersive Web SDK

Meta has directly integrated agentic AI into its open-source WebXR SDK (Immersive Web SDK) to automate code generation, testing, and validation. This development aims to reduce technical friction and accelerate VR web development.

Windsurf 2.0 beats VS Code and Cursor at their own game

Windsurf 2.0 is shifting AI-native Integrated Development Environments (IDEs) from simple editor sidecars to autonomous multi-agent mission control systems. This new version features integrated Kanban task management and persistent workspace context, enhancing developer productivity.

Vibe Coding Security Weekly — May 11, 2026: RedAccess Finds 380K Exposed Apps, TrustFall Hits AI Agents, Replit Ships Security Agent | VibeEval

RedAccess has disclosed 380,000 exposed vibe-coded applications, with approximately 5,000 leaking sensitive data. This comes alongside the simultaneous disclosure of the TrustFall prompt-injection attack and the CLI-Anything agent-integration-layer backdoor, significantly escalating the supply-chain risk for AI coding agents.

GitHub Copilot Deprecates Grok Code Fast 1 Model Effective May 15 | BotBeat

GitHub is accelerating the deprecation of its Grok Code Fast 1 model, with the change effective May 15, 2026, due to provider discontinuation. Users are required to migrate their workflows to alternative models before this date.

The ReSharper 2026.2 Early Access Program Begins: Bringing More AI Agents into Visual Studio | The .NET Tools Blog

JetBrains has launched the ReSharper 2026.2 Early Access Program, introducing Junie, its first AI agent that supports the open Agent Client Protocol (ACP). This integration aims to enable vendor-neutral AI capabilities within Visual Studio.

Sakana AI's RL Conductor Orchestrates GPT-5, Claude Sonnet 4 - Digital Mind News

Sakana AI's new RL Conductor utilizes reinforcement learning to dynamically orchestrate multiple frontier LLMs, including GPT-5, Claude Sonnet 4, and Gemini 2.5 Pro. This system has demonstrated superior performance over individual models in reasoning and coding benchmarks.

Cloud & platform providers

AWS Launches OpenAI Integration and Quick Desktop AI Tool - Digital Mind News

Amazon Web Services (AWS) has integrated OpenAI's GPT-4o and o1 models into its Bedrock service, following a restructured Microsoft partnership with OpenAI. This signifies a move towards multi-platform model availability, expanding access beyond previous cloud exclusivity.

Amazon Bedrock Guardrails supports cross-account safeguards with centralized control and management | Amazon Web Services

Amazon Bedrock Guardrails now offers cross-account centralized enforcement for AI safety controls across entire AWS organizations. This new capability simplifies governance and ensures consistent application of responsible AI policies for large enterprises.

The new AI-powered Google Finance is expanding to Europe.

Google is expanding its AI-powered Finance platform to Europe, offering local language support and advanced features. These include AI-driven research, live earnings tracking, and real-time market intelligence tailored for European users.

AI policy, regulation & governance

OpenAI offers EU access to new AI hacking model – POLITICO

OpenAI is offering EU regulators access to a cybersecurity-focused variant of GPT-5.5. This initiative aims to help European authorities identify and defend against AI-enabled cyberattacks, a move that contrasts with Anthropic's more restricted Mythos model.

Meta Faces Internal Backlash Over AI Employee Monitoring Program

Meta's employee monitoring system, implemented for AI training, has reportedly sparked internal privacy backlash. This incident is occurring amidst broader AI restructuring within the company and growing concerns about job automation.

Proposed State Privacy and AI Law Update: May 11, 2026 | Privacy + Cyber + AI

Several US state legislatures have advanced or enacted privacy and AI legislation in May 2026. Notable developments include Connecticut's data broker registration law (SB 4), Iowa's signed chatbot bill (SF 2417), and Colorado's repeal-and-replace of its AI Act (SB 189), along with numerous other bills addressing AI in health, training data transparency, and surveillance.

AI labs should pass safety review to get US government contracts, group says

An advocacy group, Americans for Responsible Innovation, is calling for mandatory safety vetting of frontier AI models as a prerequisite for US government contracts. The group cites national security risks, including the potential for AI capabilities to facilitate cyberattacks, as the basis for their proposal.

DeSantis signs AI data center regulations into law - Orlando Weekly

Florida Governor Ron DeSantis has signed SB 484 into law, which prohibits ratepayers from subsidizing the infrastructure costs of large AI data centers. The new regulation mandates that tariffs must shift the full operational expenses to the data center operators themselves.

Australia launches national AI platform 'AI.gov.au' | Digital Watch Observatory

Australia has launched 'AI.gov.au', a national AI guidance platform aimed at assisting businesses and not-for-profit organizations in safely adopting AI technologies. The platform is aligned with the National AI Plan and the AI Safety Institute's objectives.

Australia Has No AI Act. It Does Not Need One to Fine You A$100 Million.

Australia's existing Privacy Act and Consumer Law establish binding AI liability without the need for a standalone AI Act. Penalties for breaches, effective December 10, 2026, can reach up to A$100 million, primarily enforced by the Office of the Australian Information Commissioner (OAIC) and the Australian Consumer and Competition Commission (ACCC).

Australian Regulator’s AI Risk Warning To Financial Institutions – SMBtech

The Australian Prudential Regulation Authority (APRA) has issued a warning stating that Australian and New Zealand financial institutions lack adequate governance frameworks to keep pace with their rapid AI adoption. APRA specifically highlighted deficiencies in identity and access management for non-human actors and indicated an intent to pursue stronger supervisory action if risks are not managed proportionately.

APRA and ASIC Sound the AI Alarm for Boards and Executives

The Australian Prudential Regulation Authority (APRA) and the Australian Securities and Investments Commission (ASIC) have issued coordinated guidance, explicitly threatening enforcement action for inadequate AI governance, risk management, and cyber resilience. This signals a unified regulatory stance requiring boards and executives to prioritize these areas.

Industry & market moves

Google to allow AI use in engineering interviews, signalling hiring shift - Storyboard18

Google is implementing a pilot program to permit engineering candidates to use Gemini during coding interviews. This signals a strategic shift in Google's hiring practices, focusing on assessing AI-assisted collaboration skills rather than relying solely on unaided problem-solving.

Broadcom and Meta Announce Multi-Year Partnership to Scale Custom AI Silicon - The AI Software Report

Broadcom and Meta have formalized a multi-generational partnership to significantly scale custom AI silicon infrastructure through 2029. This collaboration will leverage 2nm accelerators to achieve multi-gigawatt scale, strengthening Meta's proprietary XPU platform and MTIA chips.

Meta, NVIDIA announce long-term AI infrastructure

Meta and NVIDIA have formalized a long-term strategic partnership to deploy AI-optimized data center infrastructure at scale. The collaboration includes leveraging NVIDIA's Spectrum-X networking and Vera Rubin platform, aiming to enhance Meta's AI capabilities.

Krutrim Pivots to Cloud as India's AI Unicorn Faces Model - Digital Mind News

Krutrim, India's first AI unicorn, is abandoning large-language-model development to concentrate on cloud infrastructure. This strategic pivot suggests that venture-scale startups may struggle to compete with established, well-funded incumbents in the foundational AI space.

Google Cloud Commits $750 Million to Accelerate Partners' Agentic AI Development - Channel Impact

Google Cloud has committed $750 million to a new fund aimed at accelerating agentic AI development across its ecosystem of 120,000 partners, including major integrators like Accenture, Capgemini, and Deloitte.

ELECTRA AI and Iron Horse Acquisition II Corp. (Nasdaq:

ELECTRA AI, an AI-driven battery intelligence company, has announced a definitive merger agreement with SPAC Iron Horse Acquisition II Corp. This merger aims to make ELECTRA AI the first publicly traded pure-play AI Battery Intelligence company on Nasdaq, with a valuation over $250M.

Brookfield to Invest $500 Million in Strategic Partnership with OpenAI

Brookfield has announced a $500 million investment in The OpenAI Deployment Company, OpenAI's new AI deployment platform. This strategic partnership aims to accelerate the enterprise-scale adoption of AI across various industrial and services businesses.

Nscale Secures $790 Million in Financing to Support AI Infrastructure Buildout in Norway | Press Release | Nscale

Nscale has secured $790 million in committed financing from major Nordic banks to expand its Narvik AI data center by 115MW. This marks a continued momentum in Norway's largest AI infrastructure investment, supporting the growing demand for high-performance computing.

Rialtic and Exponential AI Announce Strategic Merger to Transform ...

Rialtic and Exponential AI, two prominent healthcare AI companies, have announced a strategic merger. This consolidation aims to combine Rialtic's payment accuracy rules-engine with Exponential AI's real-time decision agents, creating a unified platform for comprehensive claim lifecycle management across payers and providers.

German AI robotics firm Sereact raises $110 million

German AI robotics firm Sereact has secured $110 million in Series B funding, intended to scale its Cortex 2.0 robotic brain and facilitate expansion into the United States. The company currently has over 200 systems deployed across Europe.

Korea's biggest manufacturers back Config, the TSMC of robot data | TechCrunch

Seoul-based robotics data platform Config has secured a $27 million seed round of funding from South Korea's largest manufacturers. This investment positions Config as a critical data infrastructure layer for robot AI, similar to TSMC's role in chip manufacturing.

MoonPay Acquires Dawn Labs, Launches AI Trading Agent Dawn CLI

MoonPay has acquired AI research lab Dawn Labs, a move designed to advance its AI-native infrastructure strategy for financial services and trading automation. Alongside the acquisition, MoonPay launched the AI Trading Agent Dawn CLI, leveraging the newly acquired expertise.

AI product & feature launches

This Google Meet feature is the first AI tool I actually want to use

Google Meet's new live speech-to-speech translation feature utilizes a unified end-to-end model to provide real-time dubbing with minimal latency. This technology preserves the speaker's emotion and cadence, delivering a practical mobile AI feature directly within the widely used application.

Writer Launches Autonomous AI Agents as Job Displacement Debate Heats - Digital Mind News

Writer has launched autonomous AI agents equipped with event-based triggers that monitor business applications and execute workflows without human intervention. This marks a significant escalation in enterprise automation, fueling the debate around AI's impact on job displacement.

Red Hat targets enterprise deployment with new version of its AI platform - SiliconANGLE

Red Hat AI 3.4 has been released, introducing model-as-a-service governance, distributed inferencing, and agent management features. These additions are designed to help enterprises transition AI initiatives from experimental stages to full production deployments.

AnySearch Launches as Search Infrastructure Built for AI Agents

AnySearch has launched as a dedicated search infrastructure for AI agents, designed to aggregate vertical data sources such as finance, legal, academic, and cybersecurity via a unified API. It differentiates itself from traditional search engines by focusing on authenticated professional systems rather than public web content.

Hermes Agent Just Took OpenClaw's OpenRouter Top Spot

Hermes Agent, an open-source runtime with persistent memory and self-improvement capabilities, has for the first time surpassed OpenClaw on OpenRouter's daily token rankings. This indicates a growing developer preference for agent depth and cross-session continuity over platform breadth.

PYMNTS | AI Agents Are Coming for Government’s Hold Music

Government agencies are now able to deploy agentic AI solutions while maintaining data sovereignty and regulatory compliance. This addresses the existing gap between escalating citizen expectations for services and the speed at which government services can be delivered, through platforms like UiPath Automation Suite and Google Gemini Enterprise Agent Platform.

New Breakthrough in Embodied Intelligence: X-Humanoid Wise KaiWu Agent Gives Robots Real Awareness and Real Capability | FinancialContent

X-Humanoid's Wise KaiWu Agent introduces the industry's first global scene perception and dynamic spatial memory system for robots. This allows robots to maintain memory across time, recognize individual users, and perform complex multi-step physical tasks with over 98% real-world accuracy.

As Enterprise AI Outpaces Governance, Alation Closes the Gap with New AI Governance Offering

Alation has launched its AI Governance platform, designed to provide enterprises with a single system of record for AI asset inventory, compliance workflows, and regulatory evidence. This offering supports compliance with key frameworks such as the EU AI Act, NIST AI RMF, and ISO 42001.

Checkbox has expanded its AI Legal Front Door with three new capabilities: AI Agent Actions, AI Corrections, and Intelligent Status Update. These features enable legal teams to automate intake processes, capture institutional knowledge in real-time, and gain operational visibility through cycle time reporting.

Research with immediate practical relevance

Prompt Injection Attacks Now Hit AI Agents in the Wild — TechReaderDaily

Real-world prompt injection attacks are actively targeting AI agents in production, with evidence showing attackers exfiltrating credentials and commanding autonomous systems without needing to jailbreak models. This reveals a fundamental architectural vulnerability across all AI agent platforms.

Sakana AI and NVIDIA Introduce TwELL with CUDA Kernels for 20.5% Inference and 21.9% Training Speedup in LLMs - MarkTechPost

Sakana AI and NVIDIA have introduced TwELL, a new sparse CUDA kernel and training framework (arXiv:2603.23198, ICML 2026). TwELL achieves a 20.5% inference speedup and a 21.9% training speedup by exploiting unstructured activation sparsity in LLM feedforward layers through tile-wise sparse kernel design.

AI model finally learns to say ‘I don’t know’ in breakthrough to curb chatbot overconfidence - AOL

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have developed a neural network warm-up training method inspired by human brain development. This breakthrough enables AI models to recognize when they lack knowledge, significantly reducing overconfidence and hallucinations in chatbots.