The Informed Layer · Newsscan
Daily briefing

Agents proliferate; cloud deals, regulation intensify

← All issues

Anthropic / Claude ecosystem

Claude Code v2.1.128 Released with Key Improvements

Claude Code version 2.1.128 introduces over 30 improvements, including a 3x lower cache creation cost for sub-agents and .zip plugin support. The update addresses multiple bug fixes across the CLI stack, enhancing developer productivity and agent reliability. This release continues rapid iteration on Anthropic's AI coding tool.

Anthropic's Claude Managed Agents Gain Dreaming Feature

Anthropic announced a 'dreaming' capability for Claude Managed Agents that allows the AI to review and self-improve its memory between sessions. The feature is part of a broader update including outcomes tracking and multi-agent orchestration, designed to handle complex, long-running tasks with minimal human steering. Dreaming helps reduce errors and improve consistency in agent behavior over time.

Frontier model providers

OpenAI Launches ChatGPT Futures Class of 2026

OpenAI introduced ChatGPT Futures, a program celebrating and supporting AI-enabled student builders with grants and access to frontier models. The initiative aims to foster a new generation of AI developers and researchers. It includes resources and community support for participants working on innovative projects.

OpenAI Rolls Out GPT-5.5 Instant as Default ChatGPT Model

OpenAI released GPT-5.5 Instant, a new default model for ChatGPT that promises improved factual reliability, reduced hallucinations, and enhanced personalization. It leverages memory and connected services to deliver more accurate and context-aware responses. The rollout targets Plus and Pro users with increased speed and efficiency.

Google AI Search Now Quotes Reddit and Forum Sources

Google updated its AI-powered search overviews to include quotes from Reddit and other web forums, providing firsthand perspectives. The feature helps users see community-sourced advice directly in search results. This aims to improve the richness and credibility of AI-generated summaries.

AI developer tooling & infrastructure

GitHub MCP Server Adds Secret Scanning General Availability

GitHub has made secret scanning generally available for its MCP server, allowing AI coding tools to detect secrets like API keys before they are committed. The feature integrates with MCP-compatible AI assistants, enhancing security during AI-driven development. It offers pre-commit scanning to prevent credential leaks.

Atlassian MCP Updates Target AI Token Usage and Context Sharing

Atlassian released updates to its Model Context Protocol (MCP) integrations, introducing an open-source standard and CLI tools to optimize context sharing for AI agents. The improvements aim to reduce token waste and improve graph accuracy when agents interact with Atlassian tools and third-party platforms. This helps keep AI integrations cost-effective.

MetaSearchMCP Provides Unified Search Backend for AI Agents

MetaSearchMCP, an open-source model context protocol server, was introduced to unify over 20 search providers into a machine-consumable backend for AI agents. It enables AI assistants to query multiple search engines and APIs through a single MCP interface, simplifying data retrieval. The tool targets developers building agentic search applications.

Airbyte Agents Fix Data Problems for AI Agent Deployments

Airbyte launched Airbyte Agents, a pre-assembled replicated data index designed to power AI agents. The product reduces data-related failures and token usage by providing clean, consistent data pipelines. This addresses a common pain point in operationalizing AI agents.

Middleware Launches OpsAI SRE Agent for Proactive Issue Resolution

Middleware introduced OpsAI, an AI-native Site Reliability Engineering agent that auto-resolves production issues with over 80% self-resolution in live environments. The agent detects and fixes problems before they impact end users. This indicates maturing AI ops tools for DevOps.

Cloud & platform providers

AWS MCP Server Now Generally Available

Amazon Web Services has made its MCP Server generally available, providing a managed remote Model Context Protocol server for secure, authenticated AWS access from AI agents. New capabilities include IAM context keys, documentation retrieval, and run_script sandboxing, allowing agents to interact with AWS services more safely and efficiently. This infrastructure offering aims to simplify integrating AI agents into cloud management workflows.

AWS WorkSpaces Gives AI Agents Their Own Managed Desktop

Amazon Web Services launched a preview of AI agents operating within Amazon WorkSpaces desktop environments, enabling agents to interact with legacy applications without API rewrites. The feature includes full audit trails and IAM-based authentication for secure, governed automation. It aims to bring AI to enterprise workflows running older software.

Amazon SageMaker AI Integrates MLflow v3.10 for Generative AI

Amazon SageMaker AI now supports MLflow v3.10, bringing enhanced experiment tracking and model evaluation for generative AI workflows. The integration provides better observability and lifecycle management for LLM and agent development on the AWS platform. It streamlines the path from experimentation to production.

Google Cloud Launches Agent Gateway ISV Ecosystem for Security and Governance

Google Cloud unveiled Agent Gateway, a programmable data plane and partner ecosystem to govern AI agent interactions across multicloud environments. The solution introduces security and governance controls for agent-to-agent and agent-to-service communications. It aims to provide enterprises with centralized policy enforcement for AI agent ecosystems.

AWS WAF Adds AI Traffic Analysis Dashboards

Amazon Web Services introduced AI Traffic Analysis dashboards for its Web Application Firewall (WAF), extending bot control with AI-specific analytics and a growing detection catalog. The feature helps identify and block malicious AI-driven traffic patterns. It enhances protection for web applications against emerging AI-based threats.

AWS Expands OpenAI Partnership to Bring Frontier AI to Bedrock

Amazon Web Services expanded its partnership with OpenAI, making OpenAI models available on Amazon Bedrock along with enterprise controls. The collaboration includes Codex on Bedrock and Managed Agents, enabling frontier AI to be deployed at production scale with AWS's security and compliance features. This bridges OpenAI capabilities with AWS infrastructure.

Cloudflare Proposes x402 to Monetize AI Agent Traffic

Cloudflare introduced the x402 concept, a framework to monetize AI-driven automated traffic at the network level. The idea involves creating an economic layer through the x402 Foundation to sustain content creation in an era where AI agents bypass traditional web monetization. This aims to address the breakdown of web economics due to increasing AI agent traffic.

AI policy, regulation & governance

White House Preps Executive Order on AI Security in Response to Mythos

The White House is considering an executive order to require vetting of new AI models before release, prompted by concerns arising from Anthropic's Mythos vulnerabilities. The order would aim to improve national security by ensuring advanced AI systems are checked for risks. This signals a shift towards more proactive federal oversight of frontier AI development.

Trump Admin Floats Policy to Limit Contractor Input on Federal AI Use

The Trump administration is drafting policy language that would restrict contractors' ability to dictate how federal agencies use AI technology. The drafts assert government control over private-sector AI deployment in government missions. If finalized, this could reshape procurement rules and AI integration in federal operations.

Australian unions and industry groups are calling for the government to fast-track workplace AI regulations, citing fragmented laws and increasing AI adoption. There is a push for stronger rules to address bias and worker rights, while businesses caution about regulatory burden. The debate underscores the growing pressure to legislate AI in employment contexts.

Regulators Demand Stronger Agentic AI Security Controls

Government and regulatory bodies are calling for enhanced security controls and lifecycle management for agentic AI systems, warning of increased risks. The guidance highlights the need for robust governance frameworks as autonomous AI agents become more prevalent in critical sectors. The push comes amid rising cyber threats.

Industry & market moves

Anthropic and SpaceX Sign Compute Deal to Scale Claude

Anthropic has signed a deal with SpaceX to access over 300 MW of data center capacity, providing 220,000+ NVIDIA GPUs for Claude. The partnership will significantly increase usage limits for Claude Pro and Max subscribers, and is expected to boost Claude's enterprise AI service capabilities. This move addresses growing demand for Claude compute and signals a strategic infrastructure expansion for Anthropic.

Google DeepMind Takes Minority Stake in Eve Online Maker for AI Research

Google DeepMind has acquired a minority stake in CCP Games/Fenris Creations, the developer of the massive multiplayer online game Eve Online. The partnership will use a controlled offline environment to train and test AI models on complex, dynamic world simulations without impacting live players. This move signals a new direction for AI research in gaming environments.

La Banque Postale Partners with Mistral AI for Sovereign AI Deployment

French bank La Banque Postale signed a strategic partnership with Mistral AI to deploy sovereign AI on its own data center infrastructure. The collaboration will leverage Mistral's language models with enhanced security and regulatory compliance, addressing data sovereignty requirements. The move highlights growing European financial sector adoption of local AI providers.

DeepSeek in Talks for First Funding Round at $45B Valuation

Chinese AI lab DeepSeek is reportedly in discussions to raise $3–4 billion in its first external funding round, led by state-backed 'Big Fund', at a valuation of up to $50 billion. This marks a significant shift in China's AI investment landscape, with government support aiming to challenge U.S. frontier labs. Tencent is also mentioned as a potential investor.

Anthropic Strikes Massive Cloud Pact with Google

Anthropic has committed to a large-scale cloud contract with Google, signaling heavy industry concentration among a few mega-cloud providers. While financial details were not disclosed, the pact highlights the dependence of frontier AI labs on hyperscaler infrastructure. This partnership deepens the relationship between Anthropic and Google Cloud.

Greenhouse to Acquire Ezra AI Labs for Conversational Hiring

Greenhouse, a hiring software company, has entered into a definitive agreement to acquire Ezra AI Labs, which provides conversational AI for interviewing. The acquisition aims to integrate AI-driven interviews into Greenhouse's platform, scaling candidate evaluation. The deal reflects growing use of AI in HR technology.

Spirit AI Partners with Bosch to Deploy Embodied AI in China

Spirit AI, an embodied AI company, announced a partnership with Bosch to accelerate deployment of its AI models in China's industrial environments. The collaboration will provide real-world data collection and training for Spirit AI's models. This move combines AI software with Bosch's manufacturing expertise.

Boost Security Acquires SecureIQx and Korbit.ai, Raises $4M

Boost Security announced the acquisitions of SecureIQx and Korbit.ai along with an additional $4 million in funding. The moves aim to build an AI-native software development lifecycle (SDLC) security platform. This consolidation strengthens its capabilities in AI-driven code security.

EPAM and Anthropic Partner for Enterprise AI Transformation

EPAM Systems and Anthropic announced a multi-year partnership to deliver enterprise-grade, safe AI solutions using Claude and the Claude Agent SDK. The collaboration combines EPAM's engineering services with Anthropic's AI models and tools. The aim is to accelerate digital transformation for large organizations.

SAP Acquires Prior Labs for $1.16B to Build Enterprise AI Lab

SAP is acquiring German AI lab Prior Labs for $1.16 billion, aiming to create an enterprise-focused AI center around tabular foundation models. The deal signals a strategic bet on AI for structured data and may include integration with 'NemoClaw'. This move positions SAP to embed AI into its core enterprise applications.

Altara Raises $7M Seed to Accelerate Scientific R&D with Agentic AI

Altara, a startup applying agentic AI to physical sciences R&D, raised $7 million in seed funding. The company aims to speed up breakthroughs in materials and industrial processes. This reflects growing investor interest in AI for scientific discovery.

AI product & feature launches

ServiceNow Build Agent Integrates with Major AI Coding Tools

ServiceNow expanded its Build Agent to work inside every major AI coding tool with governance controls enabled by default. The offering includes free access to App Engine Management Center to ensure compliance. This move integrates AI-assisted development into enterprise-grade service management workflows.

Atlassian Opens Teamwork Graph to Third-Party AI Agents

Atlassian has opened its Teamwork Graph, containing over 150 billion connections, to third-party AI agents. This expansion allows AI assistants across different tools to leverage organizational context, enhancing collaboration and workflow awareness. The move positions Atlassian as a central hub for AI-augmented enterprise knowledge.

Horizon3.ai Debuts Safe Autonomous Cyber Defense Research

Horizon3.ai announced breakthrough research enabling safe deployment of autonomous AI for cyber defense, integrated into its NodeZero platform. The system can automatically tune critical defenses with provable stability and actions confined to a pre-approved catalog. This reduces the risk of AI-driven security tools causing disruptions.

Inworld Launches Empathetic Voice Model for AI Agents

Inworld AI released Realtime TTS-2, a frontier voice model that infers user emotion and adapts delivery in real time. The model gives AI agents contextual empathy, enabling more natural and engaging conversations. This targets applications in gaming, customer service, and virtual assistants.

Zip Launches AI Agents for Accounting to Automate Close Processes

Zip introduced AI agents for Procure-to-Pay automation, designed to help complex enterprises close their books faster and more accurately. The agents handle multi-entity accounting reconciliations and reduce manual errors. This targets finance teams looking to streamline month-end close.

Tableau Unveils Agentic Analytics Platform for Trusted Automated Actions

Tableau, a Salesforce company, announced an Agentic Analytics Platform that grounds AI agents in business knowledge to enable autonomous, trusted actions. The platform integrates data visualization with agentic workflows for decision-making. This targets business users seeking AI-driven insights and actions.

Synack's Sara AI Pentesting Agent Goes Generally Available

Synack announced the general availability of Sara AI, an autonomous penetration testing agent that combines agentic AI with human validation. The service offers continuous security validation, scaling ethical hacking efforts. It aims to improve cyber resilience for organizations.

DigiCert Launches AI Trust Tools for Agents and Content

DigiCert added AI trust tools to its DigiCert ONE platform, covering agents, models, and content. The solution verifies provenance, integrity, and authenticity across the AI lifecycle, addressing concerns about AI-generated content and agent identity. This provides a trust layer for enterprise AI deployments.

Research with immediate practical relevance

OpenAI Unveils Multipath Reliable Connection Protocol to Optimize AI Training

OpenAI and partners unveiled the Multipath Reliable Connection (MRC) protocol, a new networking standard designed to reduce congestion and failures in large-scale AI training. The protocol aims to make training more energy-efficient and faster. It is being positioned for broader industry standardization, potentially benefiting all AI infrastructure providers.

Stanford Researchers Develop Multi-Agent AI for Therapeutic Nanobody Design

Stanford's Virtual Lab, a multi-agent AI system, demonstrated the ability to computationally design nanobodies for therapeutic use, completing designs in hours instead of weeks. The system orchestrates several specialized AI agents to accelerate scientific discovery. This could speed up drug development pipelines.