The Informed Layer · Newsscan
Daily briefing

AI agents drive enterprise; regulation races to catch up

← All issues

Anthropic / Claude ecosystem

Agents for financial services and insurance

Anthropic has launched ten ready-to-run financial services agent templates, integrated with Claude Cowork, Claude Code, and Claude Managed Agents. This includes new Claude add-ins for Microsoft 365 and expanded partner connectors for data access, aiming to automate and streamline financial workflows.

Inside Claude Code Auto Mode: Anthropic’s Autonomous Coding System with Human Approval Gates - InfoQ

Anthropic has enabled autonomous multi-step coding workflows in Claude Code through its new Auto Mode. This system incorporates layered safety gates and human approval checkpoints, designed to reduce approval fatigue while maintaining control over high-risk operations.

Anthropic CEO warns ‘moment of danger’ as Mythos exposes vulnerabilities

Anthropic's Mythos model has exposed tens of thousands of unpatched software vulnerabilities, leading Anthropic's CEO to warn of a 'moment of danger'. This creates a critical 6-12 month window before Chinese AI competitors catch up.

Frontier model providers

OpenAI launches ChatGPT for Intune iOS App to power enterprise and education workflows | Artificial Intelligence News - News9live

OpenAI has launched ChatGPT for Intune, a dedicated iOS app designed for enterprise and education users, featuring Microsoft Intune integration. This app enables secure AI access within managed workplace environments.

Google DeepMind quietly launched Gemma 4 under an Apache 2.0 license

Google DeepMind has released Gemma 4, its most capable open-source model family, under the permissive Apache 2.0 license. This release includes edge-optimized variants capable of running offline on mobile and IoT devices.

Google Is Building an AI Agent That Could Be Its Answer to OpenClaw - Business Insider

Google is internally testing 'Remy,' a 24/7 personal AI agent for Gemini that is designed to autonomously take actions on users' behalf. This positions Remy as a competitive response to OpenAI's OpenClaw.

AI developer tooling & infrastructure

Opsera and Cursor Partner to Embed Autonomous AI Agents Directly Into AI SDLC Workflows for Next-Gen AI-Driven Development

Opsera's autonomous DevSecOps agents can now be directly embedded into Cursor's IDE, allowing for real-time enforcement of security, compliance, and architectural standards during code generation workflows.

Copilot Cowork: From conversation to action across skills, integrations, and devices

Microsoft has expanded Copilot Cowork from desktop to iOS and Android, introducing new Work IQ skills and Dynamics 365 integrations. This enables delegated task execution across various devices, enhancing productivity and seamless workflow management.

Microsoft fixes VS Code after Copilot credited human code • The Register

Microsoft has reversed an unintended feature in VS Code that automatically credited Copilot for human-written code. The company fixed this bug in the upcoming 1.119 release after receiving developer backlash.

AI-BOMs replace SBOMs as way to track AI agents and bots • The Register

Cisco has open-sourced its AI-BOM (Bill of Materials) and Model Provenance Kit tools. These tools are designed to provide supply-chain visibility into AI models, datasets, agents, and their dependencies within enterprise environments, addressing 'shadow AI' risks.

Cloud & platform providers

Google’s Agentic Data Cloud to power ‘systems of action’ | Computer Weekly

Google has launched the Agentic Data Cloud platform, featuring over 80 updates designed to enable autonomous agents to securely access corporate data. The platform manages semantic context, governance, and cross-cloud interoperability, facilitating 'systems of action'.

AWS Introduces Generative AI Model Agility Solution

AWS has released a new framework called the Generative AI Model Agility Solution. This solution is designed to decouple AI applications from specific Large Language Model (LLM) backends, allowing enterprises to swap models without vendor lock-in.

Google Cloud Introduces Event-Driven Webhooks for AI Agents

Google Cloud has introduced Event-Driven Webhooks for AI Agents, shifting AI agents from pull-based to push-based event-driven architecture. This enables real-time reactivity to external triggers across enterprise tech stacks.

Cloudflare launches Flagship — Edge-Native Feature Flag Service

Cloudflare has launched Flagship, a native edge-based feature flag service built on OpenFeature. Flagship eliminates external service round-trips and enables sub-millisecond flag evaluation in Workers.

Gemini API File Search is now multimodal

Gemini API File Search now supports multimodal data, including images and text, along with custom metadata filtering and page-level citations. This update aims to enable more efficient and verifiable retrieval-augmented generation systems.

Gemini in Google Docs update addresses repetitive instructions

Gemini in Google Docs now supports persistent instructions, allowing users to set a preferred tone, style, and formatting once, and have it applied across all documents. This update aims to reduce repetitive command input.

AI policy, regulation & governance

AI oversight: Trump admin. will test Google, Microsoft and xAI models

The U.S. government has secured formal pre-deployment testing access to frontier AI models from major vendors including Google, Microsoft, and xAI, under the oversight of the Commerce Department's Center for AI Standards and Innovation (CAISI).

Major publishers and bestselling author Scott Turow have filed the first unified class action lawsuit against Meta, alleging willful copyright infringement. The lawsuit claims Meta used millions of copyrighted books and articles without permission to train its Llama AI model.

Advancing AI evaluation with the Center for AI Standards (US) and Innovation and the AI Security Institute (UK) - Microsoft On the Issues

Microsoft is engaging in collaborative partnerships with both the US and UK government AI institutes (CAISI and AISI) to conduct rigorous national security and safety testing of frontier AI models. This involves sharing evaluation methodologies to ensure safe development.

One in four MCP servers opens AI agent security to code execution risk - Help Net Security

Noma Security's research reveals systematic code execution and data leakage risks in Model Context Protocol (MCP) servers and Skills. The analysis identified five named attack patterns (ContextCrush, ForcedLeak, DockerDash, Replit database deletion, Amazon Q hijacking) already observed in production environments.

Colorado lawmakers advance rewrite of 2024 law to regulate artificial intelligence - Colorado Politics

Colorado lawmakers are advancing a comprehensive rewrite of the state's 2024 AI law. The new legislation aims to address implementation concerns while maintaining protections against algorithmic discrimination.

ADIA Launches Independent Review Of Privacy Code As AI Advancements Pose 'New Risks' - B&T

The Australian Data and Insights Association (ADIA) has launched an independent review of its Privacy (Market and Social Research) Code. The review aims to address new risks posed by AI advancements and technological changes, updating Australia's only APP-registered code.

Australia–Japan Commit to Partnership to Meet Cyber Security Challenges, Strengthen Cyber Defences

Australia and Japan have established a formal Strategic Cyber Partnership to strengthen regional cyber defences. This collaboration includes a focus on critical technology security, specifically mentioning artificial intelligence.

Meta Fights EU Order on Rival AI Chatbots Access for WhatsApp

EU antitrust authorities are moving towards an interim order that would force Meta to allow rival AI chatbots free access to WhatsApp. This follows an ongoing investigation into potential market abuse by Meta.

Survey: More than half of federal agencies now planning agentic AI pilots - Nextgov/FCW

A new survey reveals that 53% of U.S. federal agencies are planning agentic AI pilots. However, governance frameworks are lagging, with only 31% having implemented oversight policies despite 77% acknowledging their essentiality.

Industry & market moves

PYMNTS | FIS and Anthropic Collaborate to Enable Agent-First Banks

FIS and Anthropic are collaborating to embed engineers to co-develop AI agents specifically for bank operations. Their first joint project, the Financial Crimes AI Agent, has already been deployed at BMO and Amalgamated Bank.

After Pushback, Amazon Rolls Out Claude Code, Codex to All Employees - Business Insider

Amazon has formally expanded access to Anthropic's Claude Code and OpenAI's Codex to all corporate employees. This move resolves previous restrictions that had frustrated engineers who preferred these tools over Amazon's in-house AI coding assistant, Kiro.

OpenAI and PwC collaborate to reimagine the office of the CFO

OpenAI and PwC are collaborating to deploy AI agents across enterprise finance workflows, with initial focus on procurement and plans to expand to treasury, tax, and accounting. This leverages OpenAI's internal learnings from its own finance organization.

Google DeepMind workers in UK vote to unionize amid deal with US military

Google DeepMind workers in the UK have voted to unionize, marking the first such action at a major frontier AI laboratory. This move was primarily driven by employee concerns regarding military contracts and the ethical use of AI systems.

NVIDIA and ServiceNow Partner on New Autonomous AI Agents for Enterprises | NVIDIA Blog

NVIDIA and ServiceNow announced Project Arc, a self-evolving autonomous desktop agent with built-in governance and security controls for enterprise knowledge workers. This project is powered by NVIDIA accelerated computing and ServiceNow's AI platform.

ZyG Announces $60M Series A Round Led by Accel

ZyG, a company founded by ironSource veterans and AI experts, has successfully raised a $60 million Series A funding round led by Accel. The funding will be used to build an agentic operating system for eCommerce scaling.

Anthropic and OpenAI are both launching joint ventures for ...

Anthropic and OpenAI are both launching parallel joint ventures with major alternative asset managers. These ventures aim to deploy enterprise AI services and capture preferred portfolio company access, with Anthropic's venture valued at $1.5 billion and OpenAI's at $10 billion.

Gloo Completes Acquisition of EnterpriseMarketdesk (EMD) Bringing AI-enabled Workday Services and Expertise to the Gloo Platform

Gloo has completed its acquisition of EnterpriseMarketdesk (EMD), a Workday Services Partner. This acquisition expands Gloo's AI-enabled services for nonprofit and mid-market organizations, enhancing its platform with specialized expertise.

CopilotKit's AG-UI open standard for AI agent interaction has been adopted by Google, Microsoft, Amazon, and Oracle, and the company is expanding its Israeli development team following the funding close.

CopilotKit's AG-UI open standard for AI agent interaction has been adopted by major tech companies including Google, Microsoft, Amazon, and Oracle. The company is also expanding its Israeli development team following a significant funding close of $27 million in Series A and Seed rounds.

RadixArk Launches with $100 Million in Seed Funding Led by Accel to Grow SGLang and Democratize Frontier AI Infrastructure

RadixArk has launched as a new AI infrastructure company, securing $100 million in seed funding. Led by Accel, the funding will be used to democratize frontier AI infrastructure by commercializing SGLang and building end-to-end training and inference platforms.

ServiceNow expands AI agent governance through deeper integration with Microsoft

ServiceNow and Microsoft have deepened their integration to enable AI agent governance across the Microsoft 365 ecosystem. ServiceNow AI Specialists will also be available in the Microsoft Agent 365 Marketplace later in 2026.

AI product & feature launches

Meta Is Using AI Bone Structure Analysis To Detect And Remove Underaged Users

Meta is deploying an AI-powered visual analysis system that examines bone structure and other physical cues to detect and remove underage users from Facebook and Instagram. This expands Meta's enforcement mechanisms beyond traditional textual content analysis.

Guardrail Technologies Launches Traffic Light for Code & AI™; First Security Technology to Verify & Secure AI Code and the People Creating It | VentureBeat

Guardrail Technologies has launched Traffic Light for Code & AI™, the first security technology designed to verify and secure AI-generated code in real-time. It provides clear traffic light signals (green/amber/red) to detect zero-day threats and confirm vendor trustworthiness before code deployment.

Operant AI Launches Endpoint Protector: Securing Shadow AI,

Operant AI has launched Endpoint Protector, the first enterprise solution to discover, detect, and defend against shadow AI, coding agents, and Model Context Protocol (MCP) threats directly at the endpoint in real-time.

Dun & Bradstreet Brings Risk & Compliance Workflows to Anthropic's Claude

Dun & Bradstreet's verified business identity data is now integrated directly into Claude via a Model Context Protocol (MCP) server. This enables enterprises to automate KYC/KYB workflows with auditable, governance-ready risk logic.

Think 2026: IBM Delivers the Blueprint for the AI Operating Model as the AI Divide Widens

IBM has delivered a comprehensive blueprint for an AI operating model, integrating agent orchestration, real-time data foundations, intelligent operations, and sovereign hybrid cloud management. This aims to address the widening divide in enterprise AI adoption.

ServiceNow expands AI Control Tower to discover, observe, govern, secure, and measure AI deployed across any system in the enterprise

ServiceNow has expanded its AI Control Tower with unified governance capabilities across discovery, observation, governance, security, and measurement dimensions. It now integrates with AWS, Google Cloud, Azure, SAP, Oracle, and includes Traceloop acquisition for deep observability.

Platformr Launches Optics AI Suite, Bringing Natural Language Intelligence to AWS Organizations

Platformr's Optics Discovery, part of its new Optics AI Suite, brings natural language AI querying to AWS operations. This allows non-technical teams to access security, cost, and compliance insights without needing specialized scripting knowledge.

Causal Dynamics Lab launches Cielara Code for AI coding

Causal Dynamics Lab has launched Cielara Code, a new AI coding tool featuring a causal dependency graph. The tool is reported to outperform Claude Code and Codex in file localization accuracy while also reducing compute costs by 30–40%.

OpenAI opens up ChatGPT ads manager in U.S.

OpenAI has expanded its self-serve ChatGPT ads manager to all U.S. advertisers, dropping the minimum spend requirement. The company also outlined its roadmap for future features, including third-party measurement and CPA bidding.

Think 2026: IBM Makes Digital Sovereignty Operational with General Availability of IBM Sovereign Core

IBM has launched IBM Sovereign Core, a unified software platform that operationalizes digital sovereignty. This platform embeds governance, compliance, and control into AI systems, allowing organizations to maintain authority over their data, operations, and models.

Okta warns Australian firms on AI agent security gap

Okta has released 'Okta for AI Agents' to help organizations manage AI agent identities and close security gaps, as AI agent adoption in Australian firms is outpacing security readiness.

Pinecone Launches First Serverless Region in Asia with New Singapore Cloud Region, Bringing the Knowledge Infrastructure for AI to the Asia-Pacific Market

Pinecone has launched its first serverless region in Asia (Singapore), alongside the Nexus knowledge engine and KnowQL query language. This development aims to reduce token usage by 90% and achieve over 90% task completion rates for AI agents.

Research with immediate practical relevance

Hong Kong puts its own spin on DeepSeek with China-chip AI push abroad | South China Morning Post

The Hong Kong government-backed Generative AI Research and Development Centre (HKGAI) has unveiled the HKGAI-V3 model, an AI model specifically optimized for Chinese-made chips. This is part of a sovereign AI strategy aimed at supporting secure local deployment and overseas expansion.

Cleveland Clinic, RIKEN, and IBM Model a 12,635-Atom Protein – the Largest Known to Be Simulated with Quantum Computers

Cleveland Clinic, RIKEN, and IBM have successfully simulated a 12,635-atom protein complex, marking the largest known biologically meaningful protein modeled using quantum computers. This research used the EWF-TrimSQD algorithm and Trp-cage benchmark molecule.

AI-Powered Lab Discovers Brighter Lead-Free Nanomaterials in 12 Hours | NC State News

An AI-guided autonomous laboratory, PoLARIS (Perovskite Laboratory for Autonomous Reaction Inference and Synthesis) at NC State University, discovered brighter, lead-free nanomaterials in just 12 hours. This was achieved by running 120 experiments and learning mechanistic pathways from each result, a task that traditionally takes years.