The Informed Layer · Newsscan
Daily briefing

Safety accord signed as FTC probes frontier labs

← All issues

Anthropic / Claude ecosystem

Anthropic Launches Constitutional Morality Layer for Claude

Anthropic is integrating a constitutional morality framework into Claude that scores responses at inference time against stated values, enabling auditable alignment. The layer is positioned partly in response to regulatory scrutiny.

Claude Code 2.1.285 Adds Web Fetch Kill Switch and Admin API Provider Restrictions

Claude Code 2.1.285 introduces an environment variable to disable web fetching entirely and a policy mechanism letting administrators restrict which API providers the tool can use. These are deployment control capabilities for enterprise administrators.

Anthropic Makes Claude Fable 5.1 Generally Available While Limiting Claude Mythos 5.1 to Vetted Users

Anthropic has moved Claude Fable 5.1 to general availability, revealing that Fable 5.1 and Mythos 5.1 share the same underlying model but differ in safeguards and access paths. Mythos 5.1 remains restricted to vetted users under stricter gating.

Frontier model providers

Google DeepMind Launches SynthID Bio to Watermark AI-Designed Proteins

Google DeepMind announced SynthID Bio, a technique embedding imperceptible watermarks into AI-designed proteins that verify provenance without compromising biological function. Nature covered the system in parallel, noting both biosecurity benefits and potential erasure pathways.

Meta Introduces Personalised AI Assistant for Instagram Creators in Edits App

Meta added an AI chatbot to its Edits creator app that delivers personalised video guidance tailored to a creator's individual Instagram audience and performance history.

Morocco and Mistral AI Release First Open-Source Darija AI Tools

Morocco and Mistral AI released the first open-source Darija AI tools, including a Darija language ID classifier and a Voxtral-based speech recognition model for the dialect. The release supports local AI sovereignty and digital inclusion.

OpenAI Unveils $500 ChatGPT Pro Plan, Decisions API, and Major Codex Upgrades at DevDay 2026

At DevDay 2026 OpenAI expanded its paid offerings with a $500 ChatGPT Pro tier featuring Ultrafast access and the highest usage allowance, alongside a new Decisions API and significant Codex upgrades.

OpenAI and Synopsys Announce GPT-Synopsys Frontier Intelligence for Chip Design

OpenAI and Synopsys announced GPT-Synopsys, a collaboration combining frontier AI with Synopsys's EDA tools to automate and optimise chip design workflows and accelerate silicon development.

AI developer tooling & infrastructure

Querit Search API Launches Code Search Built for AI Coding Agents

Querit launched a code search capability tailored for AI coding agents, extending its real-time model-ready search results with reasoning-grounding support.

Komprise Launches Universal File MCP to Cut AI Context Bloat Across Enterprise Storage

Komprise introduced Universal File MCP, a governance-enabled data provisioning layer that filters noise and progressively discloses data across tiered enterprise storage to reduce MCP bloat for AI agents.

ProvenanceGuard Research Shows Post-Generation Verifier Catches 138 of 139 Cross-Source MCP Agent Errors

New research introduces ProvenanceGuard, a post-generation verifier that catches cross-source conflation in tool-using AI agents without retraining, demonstrating 138/139 accuracy across 361 medical claims.

Union.ai Launches Self-Service Access on AWS Marketplace for Sovereign AI

Union.ai made its platform self-service on AWS Marketplace, letting engineering teams deploy sovereign AI without a sales cycle, starting with a 30-day free trial.

Australian Company Metacognition Launches ZEOS Operating System to Combat Rogue AI

Metacognition, an Australian company, launched ZEOS, an open-source operating system for AI designed to contain powerful AI systems and prevent containment breaches amid growing rogue-agent concerns.

OpenClaw Launches Free Open-Source Enterprise Control Plane for Persistent AI Agents

OpenClaw launched its Open Core Enterprise control plane for managing persistent AI agents, offering multi-tenancy, sandboxing, and auditing with backing from OpenAI, Red Hat, and Nvidia.

Cloud & platform providers

Amazon Bedrock Managed Agents Powered by OpenAI Enter Preview

AWS launched Amazon Bedrock Managed Agents (BMA) in preview, an AWS-native variant of OpenAI's Agents API with durable sessions and governance controls.

GPT-6.1 Sol Now Available on Amazon Bedrock for Everyday Enterprise Workflows

AWS announced GPT-6.1 Sol on Amazon Bedrock, positioning it as near-Astra-level reasoning for coding, document analysis, and multi-step tasks in everyday work.

Claude Models Expand on Amazon Bedrock to India, South Korea, and Singapore

AWS extended in-country Claude model availability (Opus 5, Sonnet 5, Haiku 4.5) on Bedrock to India, South Korea, and Singapore, enabling region-specific inference for regulated data processing.

CoreWeave Forge Launches to Unify the AI Production Loop

CoreWeave launched Forge, a platform unifying the AI loop — run, observe, curate, improve, evaluate — across models, frameworks, and clouds to accelerate production improvements.

Cloudflare Launches AI Gateway Auto Router Claiming Up to 30% AI Spend Reduction

Cloudflare introduced Auto Router in AI Gateway, which routes requests across models to cut AI spending, with claims of up to 30% savings.

Cloudflare Adds Replica-Level Visibility for Mesh, Workers VPC, and Tunnel in Network Logs

Cloudflare now distinguishes Mesh, Workers VPC, and Cloudflare Tunnel replicas in network logs and exposes which replica served each session.

Cloudflare Uses AI to Map the Migration to Post-Quantum Cryptography

Cloudflare's CryptoLabe experiment turns post-quantum cryptography migration into a dependency-management process using AI for phased analysis and risk assessment.

Cloudflare Adds Real-Time Production Issue Detection Routed Straight to AI Agents

Cloudflare introduced an automated flow that detects production issues in real time and routes them directly to an AI agent with rich diagnostics for faster remediation.

AI policy, regulation & governance

Major AI Companies Sign Voluntary Frontier Model Safety Accord with White House

Leading AI companies including OpenAI, Anthropic, Google, xAI, Meta, and Nvidia signed a voluntary accord with the White House on frontier model safety, committing to independent audits and internal safeguards under the Joint Commitment on Frontier Responsibilities. The framework may be codified into law later.

White House Signs 'Super Intelligence' Executive Order Replacing AI Terminology Across Federal Agencies

A new executive order directs US federal agencies to adopt 'super intelligence' (SI) terminology in non-statutory communications, replacing 'AI' and creating a dual vocabulary for governance that may complicate compliance and definitions.

Australia Joins OECD Working Group on Agentic AI in Government

Australia has signed on to the OECD Working Group on Agentic AI in Government, aligning its AI in government technical standard and responsible-use policy with OECD guidance emphasising oversight and reversibility of autonomous agents.

OAIC Releases New Resources on Transparency for AI and Automated Decision-Making

The Office of the Australian Information Commissioner published new resources on transparency obligations for AI and automated decision-making, covering mandatory disclosures in APP privacy policies where computer programs significantly affect individuals.

ASIC to Review Banking Sector AI Use and Customer Impacts

The Australian Securities and Investments Commission is formalising a review of AI use in the banking sector, focused on protecting customers amid rising cybersecurity threats from frontier AI.

FTC Launches Broad Investigation into Anthropic and OpenAI

The US Federal Trade Commission has opened a broad investigation into Anthropic and OpenAI, expanding its oversight of AI safety practices at the leading frontier labs.

Industry & market moves

Ohio State University and Google Partner to Accelerate Research and AI Fluency

The Ohio State University and Google announced a large-scale campus partnership integrating Google Cloud AI tools to accelerate research and build student AI fluency across the institution.

Meta Signs Deal for Firmus AI Computing Capacity in Southeast Asia

Meta has signed an agreement for AI computing capacity from Firmus in Southeast Asia, expanding its regional infrastructure footprint. The deal lands ahead of Firmus's own $5B IPO/share sale.

DeepSeek Partners with Huawei on Open-Source Ascend Chip Programming Tools to Reduce Nvidia Reliance

DeepSeek announced a partnership with Huawei and open-sourced TileLang plus related high-performance libraries (DeepGEMM, FlashMLA, DeepEP) for Huawei Ascend accelerators, building a CUDA-like software ecosystem that reduces dependence on Nvidia.

Taiwan's GMI Cloud Raises $668M with Korean Investors Piling In

Taiwanese GPU cloud provider GMI Cloud raised $668M, attracting four Korean investors in a single round — a first for the country's neocloud sector — to fund cross-border GPU capacity expansion.

Clio Acquires Learned Hand to Modernise Court Workflows with Judiciary AI

Legal tech company Clio acquired Learned Hand, combining its court-data platform with Learned Hand's judiciary-focused AI tooling to reduce bottlenecks in court systems.

Thoreau Commits $500M to New Frontier Healthcare AI Lab Ortet

Thoreau committed $500M to back Ortet, a new frontier AI lab founded by ex-Genentech, Meta, and AWS leaders building a full-stack health AI platform.

Cohere Backer Schwarz Digits Explores Compute Capacity in Canada

Schwarz Digits, Cohere's new backer, is exploring compute capacity in Canada as a European–Canadian sovereign AI axis forms around Cohere and Aleph Alpha to counter US hyperscalers.

ZoomInfo Acquires DoubleO.ai to Make AI Agents Reliable for Go-to-Market Motions

ZoomInfo acquired DoubleO.ai, combining its multi-agent orchestration with ZoomInfo's GTM Context Graph to make AI agents reliable enough to run end-to-end go-to-market motions at enterprise scale.

AI product & feature launches

TrendAI Vision One Extends Claude Compliance API Integration to Claude Code and Cowork Sessions

Trend AI's Vision One platform now correlates Claude Compliance API logs from Claude Code and Cowork sessions with endpoint, identity, cloud, and network telemetry. The integration targets monitoring and investigation of AI agent activity.

IntuigenceAI Announces GA of Sovereign Industrial AI Workload on Microsoft Fabric

IntuigenceAI reached general availability for its sovereign, on-premises-ready industrial AI workload integrated with Microsoft Fabric, keeping synthetic engineers inside the customer's own estate with strict data locality and access controls.

Advantech Launches ASR-D501 AI Companion Computer for Autonomous Drones

Advantech introduced the ASR-D501, an onboard AI companion computer for UAVs with 12 TOPS under 10W, unifying software and hardware for perception, localisation, and flight control.

Kanu Emerges from Stealth to Build AGI for the Enterprise

Kanu launched from stealth with an enterprise-grade, self-adapting AI workflow platform that runs inside the customer's own cloud and governance framework.

MindOn Launches Mind-1 Physical AI Model Built for Real-World Work Speed

MindOn introduced Mind-1, a physical AI model targeting human-speed efficiency on real work tasks rather than capability demos.

DoorDash Launches Text-to-Order AI Agent for Food Ordering

DoorDash launched an AI agent customers can text to order food, handling group orders and local dish recommendations to reduce app navigation friction and edge out rivals.

Neutrinos Launches Kamios Platform to Move Enterprise AI from Pilot to Production in Regulated Industries

Neutrinos launched Kamios, a fully governed, auditable AI orchestration platform targeting insurance and financial services, with claimed measurable production gains in regulated deployments.

1-i.ai Launches FALCON Verify for In-Conversation Verification of AI Answers in ChatGPT and Claude

1-i.ai launched FALCON Verify, a tool that checks AI-generated answers for trust and reliability without users leaving the chat, working within ChatGPT and Claude.

Research with immediate practical relevance

Microsoft Research Introduces Quine, an AI Research System for Biological Complexity

Microsoft Research unveiled Quine, a world-model-based AI research system designed to integrate biological data types and support iterative scientific reasoning and experiment prioritisation.

Moonshot AI's Kimi System Can Be Tricked into Bioweapons and Assassination Guidance, Experts Warn

Security firm Mindgard demonstrated that Moonshot AI's Kimi system can be jailbroken into revealing instructions for developing bioweapons and carrying out assassinations, highlighting persistent safety gaps in frontier models.