The Informed Layer · Newsscan
Daily briefing

MCP ecosystem explodes; global AI rules crystallize

← All issues

Anthropic / Claude ecosystem

Anthropic introduces Claude for Small Business with 15 integrated workflows

Anthropic launched Claude for Small Business, a turnkey AI offering that integrates Claude across 15 common workflows—such as payroll, invoicing, and campaign management—by toggling on connectors for popular business tools. The product targets the large, underserved small-business segment, providing AI assistance directly inside existing applications.

Anthropic expanded Claude for the legal industry, releasing more than 20 MCP connectors and 12 practice-area plugins that integrate with law-firm tools like Thomson Reuters CoCounsel. The bi-directional integrations enable fiduciary-grade AI workflows for legal research, document drafting, and client management.

Anthropic rolls out fast mode for Claude Opus 4.7 on API and Claude Code

Anthropic introduced a fast mode for Claude Opus 4.7, increasing throughput while preserving capabilities. Available through the API and Claude Code, the mode comes at a higher price and with gated access, targeting enterprises needing higher speed for complex tasks.

Frontier model providers

Google DeepMind introduces AI‑enabled mouse pointer powered by Gemini

Google DeepMind unveiled an AI-enabled cursor that understands the visual and semantic context around the pointer, allowing users to perform cross-application tasks through gestures, speech, and context. The ‘Magic Pointer’ is part of a broader Google announcement that also includes the Googlebook laptop and Gemini Intelligence suite.

Meta launches Incognito Chat on WhatsApp, a fully private AI mode

Meta introduced Incognito Chat for WhatsApp and the Meta AI app, where AI conversations are processed in a trusted execution environment, not saved, and not used for model training. Meta claims that not even the company can read the messages.

Meta AI app now has “live AI” with real‑time voice and camera input

The Meta AI app gained a live AI feature that responds in real‑time to voice and camera input, similar to Google Gemini Live. Powered by Muse Spark, it enables natural multimodal conversations and extends Meta’s AI capabilities across its ecosystem.

DeepSeek adds chat history search, moving toward workplace‑ready tools

DeepSeek introduced a beta chat‑history search feature, letting users quickly locate past conversations. The move signals DeepSeek’s push toward features that improve productivity and enterprise workflow integration.

DeepSeek taps Huawei chips, signaling China’s AI pivot from Nvidia

DeepSeek optimized its latest model to run on Huawei chips, marking its first major shift to homegrown hardware. The move underscores China’s effort to reduce reliance on Nvidia amid ongoing export controls and build a domestic AI supply chain.

AI developer tooling & infrastructure

Cursor Bugbot adds default, high, and custom effort levels for PR review

Cursor’s Bugbot now supports per‑repo review effort levels—Default, High, and Custom—with English‑language routing rules. The update helps teams prioritize AI‑assisted code reviews more effectively.

GitHub Copilot code review adds severity levels and grouped comments

GitHub Copilot code review now surfaces comments with severity levels and groups them for easier triage. The improvement aims to reduce noise and help developers focus on the most important suggestions.

Cursor is now available in Microsoft Teams

Cursor’s agentic coding engine can now be invoked from Microsoft Teams, allowing developers to trigger coding tasks and PR creation through a chat interface. The integration brings AI‑assisted development into the collaboration flow.

Endor Labs expands AURI from securing code to securing AI coding agents

Endor Labs extended its AURI platform to govern and firewall AI coding agents within the software supply chain. New capabilities include agent governance and a package firewall to prevent malicious package introductions.

Shoppable launches first universal checkout MCP server for AI assistants

Shoppable introduced an MCP server that turns AI assistants into a multi‑brand storefront, supporting cross‑brand catalogs and single‑cart purchases inside chat. The server unifies checkout across multiple retailers for AI‑enabled commerce.

Docker AI Governance unlocks safe agentic autonomy at runtime

Docker announced AI Governance capabilities that provide centralized runtime control over AI agents and their MCP tool calls. The feature ensures enterprise guardrails are enforced automatically as agents operate.

Power Apps MCP server introduces closed‑loop learning for enterprise agents

Microsoft added closed‑loop learning to the Power Apps MCP server, enabling enterprise agents to automatically memorize user corrections and improve over time without additional pipelines. The feature brings production-grade continuous learning to low-code agentic workflows.

Cloud & platform providers

No significant new developments.

AI policy, regulation & governance

Cohere achieves FedRAMP High authorization, opening secure AI to federal agencies

Cohere became the first independent LLM provider to achieve FedRAMP High authorization through a partnership with Second Front/Game Warden. The inherited authorization enables federal agencies to rapidly deploy Cohere’s enterprise AI without needing a government sponsor.

US Commerce Department quietly deletes AI security testing deal with Google, Microsoft, xAI from website

The US Commerce Department removed details about a May 5 agreement to test AI models from Google, xAI, and Microsoft from its website. The pact aimed to give government scientists early access to identify security risks, but its status is now uncertain.

Spain defies Big Tech pressure with tough AI and social media rules

Spain advanced a regulatory package that imposes strict rules on high‑risk AI and social media, including algorithm transparency and an under‑16 social media ban, despite intense US tech lobbying. The rules align with broader EU digital regulation.

Canadian privacy ruling on AI training data sets a bad precedent, critics say

Canadian privacy regulators issued a ruling deeming OpenAI’s use of public data for training to be non‑compliant with consent standards, sparking criticism that it could hinder AI development. The decision constrains how companies can use publicly available data for model training.

G7 countries spell out key elements of AI software bill of materials

G7 working guidance outlines baseline components for AI software bills of materials (AIBOMs), aiming to increase transparency and trust in AI systems. The guidance is meant to help organizations track and attest to the provenance and composition of AI models.

Australian Greens secure parliamentary inquiry into AI data centres

The Australian Greens successfully moved for a Senate inquiry into the environmental, economic, and social impacts of AI data centres. The inquiry will scrutinise energy and water use amid rapid data‑centre expansion.

AUSTRAC releases risk updates, naming AI as a cross‑cutting accelerant for financial crime

Australia’s financial crimes regulator AUSTRAC published new national risk assessments that identify AI as a key enabler accelerating money laundering, terrorism financing, and proliferation financing. The update comes weeks before around 80,000 new firms enter the regulated perimeter.

Industry & market moves

xAI in talks for partnership with Europe’s Mistral AI to challenge OpenAI and Anthropic

xAI is reportedly in exploratory talks with European AI firm Mistral AI and developer‑tool provider Cursor. The potential partnership would combine Mistral’s models, Cursor’s coding environment, and xAI’s infrastructure to create a competitive alternative to OpenAI and Anthropic.

Anthropic and xAI are teaming up to stop OpenAI

In a surprising alliance, Anthropic and xAI are reportedly collaborating to counter OpenAI’s dominance, including a compute deal where Musk’s xAI will lease its Colossus‑1 cluster to Anthropic to scale Claude. The move reshapes partnerships and compute dynamics among top AI labs.

Mistral develops cybersecurity AI model for European banks

Mistral AI is building a dedicated cybersecurity AI model for European banks to detect vulnerabilities at scale. The initiative aims to offer an alternative to Anthropic’s Mythos, expanding secure AI to more institutions.

SAP and AWS enable next‑generation AI with bi‑directional zero‑copy data sharing

SAP introduced Business Data Cloud Connect for Amazon Athena, enabling bi‑directional zero‑copy data sharing with AWS Athena. The integration accelerates self‑service analytics and AI model development directly on SAP data without data duplication.

Carbon Arc acquires Neustreet to extend AI‑economy data infrastructure

Carbon Arc acquired Neustreet to strengthen its data‑ingestion and monetization capabilities for the AI economy. The deal aims to scale AI‑ready data infrastructure for enterprises.

Mind Robotics announces $400M in new funding to expand industrial robotics deployment

Mind Robotics secured a $400 million funding round led by Kleiner Perkins to scale deployments of its industrial robots. The capital will accelerate production and installation across logistics and manufacturing.

Nvidia partners with David Silver’s AI startup Ineffable Intelligence on reinforcement learning

Nvidia announced a partnership with Ineffable Intelligence, the startup founded by DeepMind alum David Silver, to scale reinforcement‑learning infrastructure for ‘superlearners’ that learn from experience. The collaboration targets next‑generation AI training paradigms.

AI product & feature launches

Aderant introduces Agent Center to automate law firm back‑office operations

Aderant launched Agent Center, a framework of AI agents for collections, appeals, and talent evaluation, integrated with its Stridyn and MADDI platforms. The automation targets law firm back‑office workflows to reduce manual overhead.

Microsoft makes Copilot easier to summon across Office apps

Microsoft updated Copilot in Word, Excel, and PowerPoint with simplified entry points and new keyboard shortcuts, aiming to boost user engagement with the AI assistant. The changes apply across Windows and Mac.

Manifold expands Manifest to 7,700 MCP servers with safety scores

Manifold Security expanded its Manifest platform to cover 7,700 MCP servers, adding per‑server Lineage and Safety Scores to help security teams assess and govern AI agent connections.

Workable launches MCP server for AI‑powered recruiting and HR workflows

Workable released a GA MCP server that lets AI assistants read and write data across the entire HR platform. Enterprise recruiters can now use AI agents to automate candidate sourcing, screening, and scheduling directly through conversational interfaces.

Everlaw releases MCP integration to connect Claude with litigation data

Everlaw introduced an MCP connector that integrates Claude with its e‑discovery and litigation platform, enabling agentic AI workflows while maintaining governance and data security. Lawyers can now query case data and generate insights through Claude.

Red Hat opens Ansible to AI agents with guarded execution

Red Hat released a technology preview of an MCP server and deterministic orchestrator for Ansible Automation Platform, allowing AI agents to execute pre‑approved playbooks. The approach ensures that agent‑driven IT automation stays within safe bounds.

BasedAI emerges from stealth with Hirebase, an AI workforce platform for businesses

BasedAI launched Hirebase, an open‑weight platform that provides persistent multi‑agent AI workers for business tasks, integrating Warden App technology. The platform targets enterprise productivity through agentic workflows across common tools.

Microsoft unveils MDASH, a multi‑model agentic security system that tops industry benchmark

Microsoft disclosed MDASH, a new multi‑model agentic security scanning harness that achieved top marks on the CyberGym benchmark and found 16 new Windows vulnerabilities with zero false positives. The system leverages multiple AI models in an agentic workflow.

Amazon ditches Rufus AI chatbot in favor of Alexa shopping agent

Amazon launched Alexa for Shopping, a personalized AI assistant integrated into the Amazon search bar and mobile app, replacing the standalone Rufus chatbot. Powered by Alexa+, the new agent offers proactive recommendations and cross‑site shopping capabilities.

Poppy debuts a proactive AI assistant to organize your digital life

Poppy launched a proactive AI assistant that manages calendar, email, messages, and other apps, suggesting actions and automating routine tasks. The app aims to act as an ambient personal assistant, reducing manual digital housekeeping.

SAP announces Joule Studio for enterprise‑scale agentic development

SAP launched Joule Studio, a new environment on the SAP Business AI Platform for building, governing, and managing AI agents and workflows at enterprise scale. It includes support for Cursor and Claude Code, enabling end‑to‑end agentic development.

Research with immediate practical relevance

No significant new developments.