The Informed Layer · Newsscan
Daily briefing

AI security and regulation advance; agents drive infra

← All issues

Anthropic / Claude ecosystem

TrendAI™ and Anthropic Advance AI-Powered Vulnerability Detection and Risk Mitigation with Claude Opus 4.7

TrendAI and Anthropic have partnered to enhance AI-accelerated vulnerability discovery and mitigation. Their collaboration leverages Claude Opus 4.7 to bridge the gap between threat identification and real-world risk reduction, integrating it with TrendAI's AESIR platform and Cyber Verification Program.

Claude Security Public Beta Launches in Claude Code on Web | gentic.news

Anthropic has launched Claude Security in public beta, directly embedding vulnerability detection and remediation capabilities into the Claude Code editor. This integration streamlines the security review process, allowing developers to address issues within their development environment rather than using separate tools.

Frontier model providers

OpenAI Expands Trusted Access Program With GPT-5.5-Cyber - Dataconomy

OpenAI is expanding access to its cybersecurity-focused AI, GPT-5.5-Cyber, through its Trusted Access program. This move contrasts with Anthropic's more restricted rollout of its Claude Mythos model, indicating differing strategies for making advanced AI security tools available.

xAI launches Grok 4.3 with improved agentic performance and lower pricing

xAI has launched Grok 4.3, an updated model that boasts improved agentic performance (ELO +321 on GDPval-AA) while simultaneously reducing inference costs by approximately 40–60% compared to its predecessor, Grok 4.20.

DeepSeek begins limited testing of vision mode · TechNode

DeepSeek is beginning limited testing of its new vision mode, marking the company's first entry into multimodal capabilities. This new mode will operate alongside its existing Fast and Expert modes.

Google Launches Gemini Embedding 2: One AI Model for Text, Images, Video, Audio, and Documents - UNDERCODE NEWS

Google has launched Gemini Embedding 2, a unified multimodal embedding model capable of processing text, images, video, audio, and documents within a single semantic space. This eliminates the need for enterprises to maintain separate systems for each media type.

AI developer tooling & infrastructure

Cursor Introduces a TypeScript SDK for Building Programmatic Coding Agents With Sandboxed Cloud VMs, Subagents, Hooks, and Token-Based Pricing - MarkTechPost

Cursor has released a public beta TypeScript SDK, shifting its focus from an interactive IDE to programmable infrastructure. This SDK allows developers to programmatically invoke coding agents, supporting sandboxed cloud execution and multi-agent workflows with token-based pricing.

ActiveState Curated Catalog Secures AI-Generated Code Across Any Development Environment

ActiveState has extended its Curated Catalog to provide tool-agnostic open-source dependency governance for all AI coding assistants, including Cursor, Claude Code, and GitLab Duo. This ensures enterprise-grade security at the point of AI code generation.

Visual Studio Code Update Adds Remote Copilot Control and Enterprise AI Restrictions

Visual Studio Code version 1.118 introduces remote Copilot CLI control across devices and new enterprise AI access restrictions. These restrictions are tied to GitHub organization authentication, enhancing control over AI usage in corporate environments.

DBmaestro MCP Server Puts Natural Language in Control of Database Pipelines - InfoQ

DBmaestro has launched its MCP server, providing AI agents with access to enterprise database DevOps workflows. This solution preserves role-based access control and audit compliance, addressing the historical challenges of automating database layers.

CVE MCP Server Turns Claude Into a Fully Capable Security Analyst With 27 Tools Across 21 APIs

An open-source CVE MCP Server has been released, integrating Claude AI with 27 security tools across 21 APIs. This enables Claude to automate vulnerability triage, eliminating the need for manual context-switching across various intelligence platforms.

Keeper Security Launches Agent Kit to Secure AI-Driven Developer Workflows

Keeper Security has launched its Agent Kit, designed to enable AI coding agents to securely retrieve secrets and manage infrastructure. This prevents the exposure of credentials in chat history or source control, enhancing security for AI-driven development.

Command Zero Launches APIs and MCP Server for Autonomous SOC & AI Security Operations

Command Zero has opened its autonomous SOC investigation engine via new APIs and an MCP server. This allows customers to integrate advanced threat hunting and remediation capabilities into their existing SOAR playbooks and AI agents.

Cloud & platform providers

What’s New in Microsoft 365 Copilot | April 2026 – Azure Feeds

Microsoft has rolled out significant feature updates for Copilot across various Microsoft 365 applications, including Excel, PowerPoint, Word, Teams, OneDrive, Outlook, and Copilot Notebooks. New additions include 'Plan mode,' Python support, multimodal capture, and a mobile app refresh.

AI policy, regulation & governance

Musk Admits Distilling OpenAI Data For His XAI—Here’s Why That’s Controversial - Forbes Africa

Elon Musk has publicly admitted under oath that xAI utilized OpenAI's technology through distillation to train its own models. This action potentially violates OpenAI's terms of service and is central to a high-profile lawsuit concerning OpenAI's shift from a nonprofit to a for-profit entity.

Italy’s antitrust authority closes probes into DeepSeek, Mistral, and Nova AI

Italy's AGCM (Autorità Garante della Concorrenza e del Mercato) has concluded investigations into DeepSeek, Mistral, and Nova AI. This makes AGCM the first European regulator to secure binding commitments from AI companies, requiring them to provide contextual hallucination warnings as a consumer protection measure.

Senator Hawley’s GUARD Act to Protect Kids from AI Chatbots Passes Committee Unanimously - Josh Hawley

The Senate Judiciary Committee has unanimously advanced bipartisan legislation known as the GUARD Act. This bill proposes criminal penalties for AI companies whose chatbots engage in sexually explicit conduct with minors or solicit self-harm.

Warner, Budd Introduce Legislation to Collect Data on AI's Impact, Guide Lawmakers

U.S. Senators Mark Warner and Ted Budd have introduced bipartisan legislation, the Workforce Transparency Act, to establish a federal framework for the Department of Labor. This framework will collect and publicly share aggregated data on AI usage and its impact across both public and private sectors.

Why You Need To Engage And Comment On The Draft South Africa National AI Policy - Telecoms, Mobile & Cable Communications - South Africa

South Africa's Department of Communications and Digital Technologies has released a comprehensive draft AI policy with a proposed three-year implementation timeline. The policy suggests new regulatory bodies and risk-based governance frameworks, notably aligned with the African Ubuntu philosophy.

China launches campaign to rectify improper AI content production - CGTN

China's Cyberspace Administration of China (CAC) has initiated a four-month campaign to curb AI-generated disinformation, malicious content, and low-grade material. The campaign will also review model registration and safety mechanisms.

White House AI Memo Hits Issues in Anthropic-Pentagon Feud - Bloomberg

The White House is currently drafting an AI policy memo that addresses vendor concentration and military chain-of-command safeguards. This memo comes amidst an ongoing dispute between the Pentagon and Anthropic regarding AI deployment.

APRA warns super funds about AI-related risk management and governance - techpartner.news

APRA's supervisory review found that Australian financial institutions' AI governance and information security practices are not keeping pace with AI adoption risks. The authority has formally warned superannuation trustees that AI governance must be treated as a prudential risk, with boards expected to maintain technical literacy and effective oversight, or face stronger supervisory action.

Industry & market moves

AWS Will Be An OEM, Just Like Google And Maybe Microsoft

Amazon Web Services (AWS) is reportedly pivoting to become a hardware OEM supplier, providing Trainium chips and racks to major AI companies like Anthropic and OpenAI. This strategy mirrors Google's approach with Broadcom and TPUs.

Singapore co-founded AI firm Featherless.ai raises $20M funding co-led by AMD Ventures, Airbus Ventures - TNGlobal

Featherless.ai, an AI firm co-founded in Singapore, has successfully raised $20M in Series A funding. This round was co-led by AMD Ventures and Airbus Ventures, with the aim of scaling open-source AI infrastructure as a neutral alternative to proprietary cloud and hardware ecosystems dominated by hyperscalers.

SKF commits to a sovereign AI future through strategic partnership with Sferical AI

SKF has announced a strategic partnership with Sferical AI, committing to a sovereign AI future. This involves securing dedicated compute capacity on Sweden's sovereign AI supercomputer, aimed at protecting proprietary industrial data while accelerating AI deployment for predictive maintenance and engineering optimization.

Palo Alto Networks To Acquire Portkey To Boost AI Security Play

Palo Alto Networks is set to acquire Portkey, an AI infrastructure startup. This acquisition aims to strengthen Palo Alto Networks' enterprise AI security capabilities, particularly in response to the rapid adoption of autonomous AI agents.

Trust3 AI and Dell Technologies Partner to Deliver Secure, Governed AI-Ready Data Lakehouse Infrastructure

Trust3 AI and Dell Technologies have partnered to integrate unified governance directly into Dell storage platforms, including ECS, ObjectScale, and PowerScale. This collaboration aims to enable secure and auditable AI workloads across hybrid environments, addressing the 'Governance Friction Gap' for agentic AI.

Agentic engineering startup JuliaHub lands $65M to automate design and testing of industrial products - SiliconANGLE

JuliaHub Inc. has secured $65M in Series B funding to scale its Dyad 3.0 platform, an agentic AI solution that automates industrial engineering design and testing. The platform aims to accelerate these processes to the speed of software development.

Datavault AI and CyberCatch Announce Signing of Binding Letter of ...

Datavault AI Inc. and CyberCatch have signed a binding letter of intent for Datavault AI to acquire CyberCatch. This all-stock acquisition aims to combine AI-driven cyber risk mitigation with quantum-resistant encryption technology to address the $240B global information security market and the emerging post-quantum security threat horizon.

Standard Intelligence raises $75M to develop efficient computer use models - SiliconANGLE

Standard Intelligence has raised $75M in funding to further develop its FDM-1, a computer use foundation model. This model is trained on video and is claimed to outperform screenshot-based alternatives with 100 times more efficient video encoding.

OpenAI Reaches 10-Gigawatt AI Capacity Milestone Years Ahead of Target - Bloomberg

OpenAI has achieved its 10-gigawatt computing capacity target years ahead of its initial 2029 deadline. This accomplishment is accelerating the company's data center expansion plans.

AI product & feature launches

AI-Designed Drugs: DeepMind's Biotech Arm Nears Crucial Clinical Trials

Isomorphic Labs, DeepMind's biotech subsidiary, is preparing to initiate clinical trials for AI-designed drugs. These drugs leverage the Nobel Prize-winning AlphaFold technology, marking a significant step towards validating AI's role in pharmaceutical discovery.

X Launches Major AI-Powered Ad Platform Overhaul in Partnership With xAI, Introduces Semantic Targeting and Redesigned Ads Manager | 📲 LatestLY

X (formerly Twitter) and xAI are collaborating to rebuild X's two-decade-old ad infrastructure. The overhaul will incorporate semantic targeting and AI-driven contextual ad delivery, aiming to boost advertiser ROI and enhance the user experience.

Daloopa Integrates with Perplexity to Bring Its Trusted Financial Data Layer into AI Research Workflows

Daloopa's financial data layer is now directly integrated into Perplexity's AI workflows through a bring-your-own-license (BYOL) model. This integration allows investment teams to conduct complex analyses using trusted, structured financial data without requiring API configuration.

IBM Releases Two Granite Speech 4.1 2B Models: Autoregressive ASR with Translation and Non-Autoregressive Editing for Fast Inference - MarkTechPost

IBM has released two new Granite Speech 4.1 2B models, demonstrating that careful architectural design can achieve competitive Automatic Speech Recognition (ASR) accuracy and fast inference (RTFx ~1820) in a 2B-parameter open model. These models feature a conformer encoder with CTC, a modality adapter, and a fine-tuned LLM.

Hippocratic AI Launches Polaris 5.0: The First Evidence- based AI for Healthcare Proven to Outperform Every Frontier Model on Critical Medical Tasks and Safety

Hippocratic AI has launched Polaris 5.0, an AI model specifically designed for healthcare. It claims to surpass frontier models from GPT, Claude, and Gemini in clinical accuracy, safety, and real-time voice performance within care settings.

SUPERAGENT AI Unveils Platform 2.0, Delivering a Unified Autonomous Workforce for Independent Insurance Agencies

SUPERAGENT AI has unveiled Platform 2.0, which consolidates five core infrastructure pillars: unified AI agents, managed telephony, automated email deliverability, real-time ACORD quoting, and enterprise security. This creates a single autonomous workforce platform for independent insurance agencies.

X announces a rebuilt ad platform powered by AI | TechCrunch

X (formerly Twitter) is rebuilding its advertising platform, incorporating AI-driven retrieval, ranking, and targeting mechanisms. This initiative aims to revitalize advertiser growth following several years of revenue decline.

Spotify introduces verified artist badges to help distinguish humans from AI | TechCrunch

Spotify has introduced a new verification badge system to help listeners differentiate between human artists and AI-generated content. This initiative comes as the platform experiences a surge in AI-generated tracks.

Clink has launched the world's first fiat agentic payment skill, enabling AI agents to autonomously pay merchants using user credit cards. This system operates within pre-defined spending limits and integrates with over 135 currencies and 100 local payment methods.

Experian Announces Agent Trust to Power Trusted AI Driven Commerce

Experian has announced 'Agent Trust,' a first-of-its-kind human-to-agent binding service designed to power secure AI-driven commerce. This service extends identity verification into autonomous transactions.

Loti AI has launched Interchange, an industry-first platform designed to manage consent, control, and compensation for the use of digital likenesses and creative works in generative AI training and generation.

Research with immediate practical relevance

Google Unveils TurboQuant: AI Models Use 6x Less Memory Without Performance Loss

Google has introduced TurboQuant, a new compression algorithm featuring PolarQuant and QJL techniques, which allows AI models to use up to six times less memory during inference without compromising performance. This innovation significantly improves efficiency for deploying AI models.

Anthropic Releases BioMysteryBench To Evaluate AI Capabilities In Bioinformatics Research

Anthropic has released BioMysteryBench, a new bioinformatics benchmark featuring 99 expert-authored questions. This benchmark includes superhuman tasks that current human experts cannot solve, aiming to address limitations in existing scientific AI evaluations.