The Informed Layer · Newsscan
Daily briefing

AI agents ship; Regulation seeks coordination

← All issues

Anthropic / Claude ecosystem

No significant new developments.

Frontier model providers

OpenAI Open-Sources Codex Security CLI to Find, Validate, and Fix Code Vulnerabilities

OpenAI has open-sourced its Codex Security CLI and TypeScript SDK, extending the Codex ecosystem into application security. This new tool integrates vulnerability scanning directly into developer workflows and CI pipelines, aiming to lower friction for developers already using OpenAI credentials.

AlphaFold 3 Fine-Tuning API Opens to Pharma Researchers

Google DeepMind has launched a fine-tuning API for AlphaFold 3, allowing pharmaceutical researchers to customize the model on their proprietary molecular datasets via Google Cloud. This moves AlphaFold 3 from a research artifact to a commercial platform for drug discovery.

DeepSeek发布并开源ng28：2230亿参数MoE模型，推理速度提升5倍

DeepSeek has open-sourced ng28, a 223 billion-parameter Mixture-of-Experts (MoE) model. It reportedly achieves 5 times faster inference and 80% lower cost than GPT-4o while matching or exceeding its benchmark performance.

AI developer tooling & infrastructure

Interactive Brokers Opened the Door to (Almost) Every AI Tool. One Name Is Missing

Interactive Brokers has removed its restriction on certified AI marketplaces and published an open Model Context Protocol (MCP) connector. This allows clients to link any MCP-compatible AI tool, such as Claude Code, Cursor, Perplexity, or Windsurf, directly to their trading accounts.

X1® Introduces X1 Search MCP Connector for Claude, Bringing

X1 Discovery, Inc. has introduced the X1 Search Model Context Protocol (MCP) Connector for Claude. This integration allows enterprises to incorporate robust search capabilities into their Claude deployments, significantly reducing AI token costs from approximately $20 to fractions of a cent per search, while keeping data on-premises.

Cloud & platform providers

Add security context to operational investigations with AWS DevOps Agent and Wiz

The AWS DevOps Agent now integrates Wiz's security graph through the Model Context Protocol (MCP) to provide crucial security context during operational incident investigations. This integration allows engineers to quickly distinguish between performance issues and genuine security incidents.

AWS Reimagines Machine Learning Certification to Align with the Generative and Agentic AI Era

AWS has updated its flagship Machine Learning certification, the AWS Certified Machine Learning Engineer – Associate (MLA-C02), to explicitly validate skills in generative AI, foundation models, and agentic AI workflows. This reflects a significant industry shift from pure model-building to model consumption and orchestration.

AI policy, regulation & governance

Frontier AI developers urge international coordination to pace automated research before capabilities outstrip control

Leading AI researchers and developers from OpenAI, Google, Anthropic, and Meta have signed an open letter, 'Pacing the Frontier,' urging international coordination mechanisms to manage the pace of automated AI development. They warn of risks if AI capabilities advance faster than safety controls.

Meta’s Zuckerberg warns against curbs on Chinese AI models, FT reports

Meta CEO Mark Zuckerberg has publicly opposed proposed U.S. bans on Chinese AI models, arguing that fostering competition rather than imposing restrictions would better serve American interests.

Beijing is shopping for autonomous laboratories

Beijing is opening bidding for lab automation equipment as part of its AI-for-science plan (2026–28). This initiative aims to develop autonomous research systems, signaling a national push into AI-driven scientific discovery.

Radical rethink for South Africa's national AI policy

South Africa's revised national AI policy is moving towards a light-touch approach, empowering individual sectors to develop their own AI strategies instead of centrally prescribing rules. This framework may also streamline proposed governance institutions.

US bans foreign-made humanoid robots, targeting China

The U.S. Federal Communications Commission (FCC) has banned imports of foreign-made humanoid and quadruped robots, specifically targeting China, which currently dominates 85% of the global humanoid robot market. The ban cites cybersecurity and national security risks.

Multiple regulators, one incident: insurers flag the compliance gap

An insurance industry submission has exposed structural fragmentation in Australia's cyber incident notification requirements, revealing that a single cyber incident could trigger overlapping obligations to six different regulators due to misaligned definitions and timeframes.

Industry & market moves

Superwhisper partners with Cohere for all-Canadian voice-to-text AI

Superwhisper has partnered with Cohere to integrate Cohere's open-source transcription model into its voice-to-text dictation platform. This collaboration creates an all-Canadian solution that runs locally, avoiding the need to send data to foreign cloud services.

Disney is ditching Microsoft's GitHub Copilot and adding OpenAI's Codex

Disney is replacing Microsoft's GitHub Copilot with OpenAI's Codex while maintaining its use of Anthropic Claude and Cursor for coding. This decision comes after internal feedback indicated that Copilot lagged in usability and adoption among Disney's developers.

Bot-detection startup Spur nabs $200M fro...

Bot-detection startup Spur Intelligence has raised $200 million in Series funding from Insight Partners. This comes as malicious bot traffic has surpassed human traffic online for the first time, highlighting an escalating challenge in cybersecurity.

LexisNexis Announces Agreement with AustLII to Advance the Rule of Law and Access to Justice in the AI Era

LexisNexis has partnered with AustLII to develop AI-enhanced legal research tools. The collaboration aims to ensure free access to trusted Australian legal information while advancing the rule of law in the AI era.

Bloomberg to Acquire Canoe Intelligence, Taking a Defining Step in Its Mission to Transform Private Markets Investing

Bloomberg is acquiring Canoe Intelligence, integrating its AI-powered private markets data automation with Bloomberg's public markets infrastructure. This aims to deliver unified analytics across the entire investment lifecycle.

Robo.ai and Abu Dhabi Enterprise Jointly Establish AI Industrial Group Alif Holding to Serve Infrastructure, Government and Industrial Sectors

Robo.ai Inc. and Abu Dhabi Enterprise have jointly established Alif Holding, an AI industrial group in Abu Dhabi. This new venture aims to serve government, infrastructure, and industrial sectors across the GCC region with integrated intelligent software and equipment platforms.

Vena to Acquire Morpheo AI, Advancing Vena AI Through Vena Omega™, the Industry’s Only Cumulative Context Engine Built for Finance

Vena Solutions is set to acquire Morpheo AI, with plans to launch Vena Omega, a cumulative context engine designed for enterprise finance AI workflows. This acquisition aims to enhance Vena's AI capabilities significantly.

Legora has completed its fifth acquisition of 2026, integrating Wexler's fact intelligence engine into its legal AI platform. This follows a recent $600 million Series D funding round for Legora.

Cyera to acquire Oasis Security for $1B to police AI agent identities

Data-security unicorn Cyera is set to acquire non-human identity specialist Oasis Security for $1 billion. This acquisition aims to bolster Cyera's capabilities in monitoring and policing AI agent identities as enterprises scale autonomous agents.

AI doctor startup Doctronic acquires Summer Health to move into pediatric care

AI-native primary care provider Doctronic has acquired Summer Health, a text-based pediatric care platform. This move expands Doctronic's clinical offerings and will leverage Summer Health's 100,000+ pediatric conversations to build specialized AI models for pediatric care.

Henry AI Raises $16.5M Series A and Launches Henry Deal

Henry AI, an AI platform specializing in automating back-office workflows for commercial real estate, has closed a $16.5 million Series A funding round. The company also launched 'Henry Deal,' a new system of record for deal lifecycle management.

AV and Applied Intuition Partner on Uncrewed Teaming

AeroVironment, Inc. and Applied Intuition have partnered to integrate collaborative autonomy software, enabling multi-platform swarm operations for U.S. military launched effects systems.

Pangram Launches New World's Best AI Detector, Expands to Images Backed by $9M From Menlo Ventures

Pangram has launched Pangram 4, a new text detection model, and a research preview of its AI image detector, both showing major accuracy improvements. This development is backed by a $9 million funding round led by Menlo Ventures.

AI product & feature launches

facebook news: Announcing the AI Glasses Impact Grant Recipients: Helping People Work, Learn, and Live More Independently

Meta has announced the recipients of its AI Glasses Impact Grant Program, supporting projects that develop practical applications of Ray-Ban Meta AI glasses. The grants focus on workforce development, accessibility, and education across various US states.

Zscaler’s 46-agent “Virtual SOC Workforce” draws paying customers ahead of general availability

Zscaler is deploying a modular fleet of 46 specialized AI agents, dubbed the "Virtual SOC Workforce," for security operations centers. The system is already being used by paying customers in limited availability, ahead of its general release.

Let Robots "Understand by Touch" the World, BeingBeyond Releases New Generation Model Being-H0.8

BeingBeyond has released Being-H0.8, a new-generation model that systematically integrates tactile modalities into large-scale embodied foundation model pre-training. This enables robots to understand touch alongside vision and action, enhancing their perception of the physical world.

Samaya AI Releases FrontierFinance, a New Public Benchmark for AI Agents in Investment Workflows; Samaya's AI System Outperforms Claude Fable 5 and GPT-5.6 Sol

Samaya AI has released FrontierFinance, a new public benchmark designed to measure AI agent performance across the entire investment workflow. The benchmark includes 220 queries and 11,543 expert-crafted rubrics, with Samaya's own AI system reportedly outperforming Claude Fable 5 and GPT-5.6 Sol.

BrowserStack Launches Test Companion, Agentic AI That Brings Complete Test Automation Into the IDE

BrowserStack has launched Test Companion, an IDE-integrated agentic AI for test automation. This tool manages the complete test lifecycle, including authoring, execution, debugging, and maintenance, moving beyond mere code generation.

MinIO Launches AIStor Memory, the Enterprise Memory

MinIO has launched AIStor Memory, an integrated enterprise memory system purpose-built for AI agents. This solution is designed to store persistent memory, workspaces, and secrets without requiring the assembly of multiple point systems.

MoonPay opens PayBox for AI agent transactions

MoonPay has launched PayBox, enabling AI agents to transact directly on the open internet through conversational interfaces like Claude and ChatGPT. It features user-controlled fund custody and per-transaction approval controls.

Research with immediate practical relevance

Like a molecular shrink ray, AI tool redesigns proteins without sacrificing function

Duke University School of Medicine researchers have developed an AI framework called 'Raygun' that enables protein redesign at evolutionary scales while preserving structure and function. This breakthrough has significant applications in gene therapy and biomedical research.