The Informed Layer · Newsscan
Daily briefing

Agent tools surge; Anthropic accuses Alibaba

← All issues

Anthropic / Claude ecosystem

Claude Fable 5 may return to subscriptions based on code strings

Internal strings in Claude Code v2.1.190 suggest Anthropic's Claude Fable 5 could return to a subscription model with a weekly quota, rather than pay-as-you-go credits.

Frontier model providers

Facebook launches AI companion app for creators

Meta is reimagining Creator Studio as a standalone AI companion app to help creators grow on Facebook, providing conversational guidance and daily priorities.

AI developer tooling & infrastructure

GitHub Copilot sees record usage after switch to pay-as-you-go pricing

Microsoft's GitHub Copilot experienced its best month ever in June 2026 after introducing usage-based pricing, driving higher adoption and revenue.

Undo adds deterministic root-cause analysis to AI coding agents

Undo has enhanced its platform with a 'control-Z' feature that provides runtime context for AI agents, enabling precise debugging across development, test, and production.

LucidLink has introduced a public beta Model Context Protocol server that allows AI agents to access distributed files across cloud, edge, and on-premises environments.

OpenRouter releases MCP server for model selection

OpenRouter has launched an MCP server providing up-to-date model data, benchmarks, pricing, and docs to help developers choose the best AI model for tasks.

CData launches Connect AI Developer Edition, Python SDK, and CLI

CData has released a free developer edition of Connect AI along with an open-source Python SDK and CLI, enabling governed access to enterprise data for AI applications.

Jaz open-source agent orchestration framework now available

A new open-source framework, Jaz, has been released to orchestrate multiple autonomous coding agents with a persistent visual board and client/server architecture.

NVIDIA releases open-source Agent Toolkit for secure enterprise AI agents

NVIDIA has introduced an open-source toolkit to help enterprises build secure, specialized AI agents that integrate into existing workflows.

Cloud & platform providers

NVIDIA and AWS collaborate to bring AI to production at scale

NVIDIA and AWS have announced a collaboration to streamline AI workload deployment, from development to production, with an end-to-end workflow.

Google Cloud expands Confidential Computing for verifiable private AI

Google Cloud has extended its Confidential Computing capabilities to enable verifiable, private AI workloads, strengthening data protection for sensitive tasks.

AI policy, regulation & governance

Reserve Bank of India proposes kill switches for bank AI models

The RBI released draft guidance requiring banks to implement kill switches and human oversight for AI models, addressing risks in financial services.

Australia launches Agentic AI Playbook for government use

The Digital Transformation Agency released a 32-page technical addendum guiding public bodies on deploying autonomous AI agents with trustworthiness and control measures.

Australian government faces whistleblower allegations over potential copyright law changes linked to AI training and data-center investments, raising policy influence concerns.

Industry & market moves

Anthropic accuses Alibaba of illicitly extracting Claude AI capabilities

Anthropic has accused Chinese tech giant Alibaba of running the largest known campaign to illicitly extract capabilities from its Claude AI models. The allegations could impact AI security policy and national competitiveness.

Two key Google Gemini researchers reportedly moving to Anthropic

Neowin reports that Gemini researchers Jonas Adler and Alexander Pritzel are planning to leave Google for Anthropic, signaling a talent drain at Google's AI labs.

NTT DATA partners with Cursor to accelerate AI software modernization

NTT DATA has partnered with Cursor to integrate the AI coding platform into its modernization services, aiming to speed up enterprise software development with AI governance.

Amazon commits $13 billion additional AI infrastructure investment in India

Amazon is investing a fresh $13 billion to expand AWS AI infrastructure in India, including data centers in Mumbai and Hyderabad, deepening its footprint.

Sail Research emerges from stealth with $80 million from Kleiner Perkins

AI infrastructure startup Sail Research has emerged from stealth with an $80 million funding round led by Kleiner Perkins, aiming to build AI systems beyond chatbots.

EXL to acquire iMerit for AI foundation model expertise

Data and analytics firm EXL announced the acquisition of iMerit for up to $310 million, strengthening its AI capabilities with foundation model training and reinforcement learning.

Code Metal acquires Signal Processing Technologies to expand provable AI platform

Code Metal has made its first acquisition, buying Signal Processing Technologies to add advanced RF capabilities and broaden its provable AI platform.

Agility Robotics to go public via $2.5 billion SPAC merger

Humanoid robotics company Agility Robotics has entered a definitive agreement to merge with Churchill Capital Corp XI, becoming the only US-listed pure-play humanoid firm.

Asteria and Pegasus Tech Ventures launch $10M AI and robotics investment fund

Asteria Corporation and Pegasus Tech Ventures have launched a $10 million fund to invest in AI and robotics startups.

Incode acquires Identiq to expand privacy-first identity infrastructure

Identity verification firm Incode has acquired Identiq, committing $100 million to privacy-first identity technology that uses on-device processing and zero-knowledge proofs.

ReSource Pro acquires Flow Specialty AI to form insurance AI factory

Insurance services company ReSource Pro has acquired Flow Specialty's AI technology and team, integrating them into a new AI Factory for AI-driven insurance solutions.

Scaled Cognition raises $100M led by Khosla to build reliable AI

Startup Scaled Cognition secured $100 million in Series A funding from Khosla Ventures to develop AI models that aims to eliminate hallucinations.

AI product & feature launches

AGIBOT's Spirit G2 humanoid robot achieves 99.98% inspection success rate

AGIBOT's Spirit G2 humanoid robots successfully completed a six-day live-streamed tablet quality inspection line at LONGCHEER with 99.98% accuracy, proving near-industrial deployment readiness.

Linux Foundation project aims to bring DNS-style trust to AI agents

The Linux Foundation announced the Agent Name Service (ANS) framework, designed to verify identity, ownership, and authenticity of AI agents before they interact.

Research with immediate practical relevance

Rough demos enable precise robot actions using DiSPo policy learning

Researchers have developed DiSPo, a diffusion-based policy learning method that achieves high-precision robot actions from coarse demonstrations, with up to fourfold real-world improvement.

Unconventional AI introduces Un-0 for image generation with coupled oscillators

A novel image generation method, Un-0, uses coupled oscillators instead of traditional neural networks, potentially offering a new approach to generative models.

Alibaba's Qwen-AgentWorld model improves agent performance without agent training

Alibaba researchers released Qwen-AgentWorld, a model that predicts environments to boost agent performance across seven benchmarks, without explicit agent training.

Lumos model tops zero-shot embodied AI benchmark

The Lumos model from an undisclosed developer has achieved top performance on a global zero-shot embodied AI benchmark, signaling rapid progress in generalist embodied agents.