The Informed Layer · Newsscan
Daily briefing

Agents surge; Florida sues OpenAI, EU tests Mythos

← All issues

Anthropic / Claude ecosystem

Anthropic launches Claude Cowork, shifting from chat to autonomous document execution

Anthropic has introduced Claude Cowork, a new feature that moves beyond conversational chat to enable autonomous execution of documents and workflows directly within the Claude environment. This marks an expansion of Claude's capabilities into agentic automation for enterprise tasks.

Claude Code v2.1.158 enables Auto Mode on AWS Bedrock, Vertex AI, and Palantir Foundry

The latest update to Claude Code (v2.1.158) introduces an 'Auto Mode' that allows fully autonomous agent behaviors inside compliant cloud virtual private clouds (VPCs) on AWS Bedrock, Google Vertex AI, and Palantir Foundry. This new feature enables enterprises to leverage Claude's agentic capabilities while maintaining data residency and security requirements.

Frontier model providers

OpenAI offers UK banks access to GPT-5.5 Cyber after Anthropic blocks Mythos access

Following Anthropic's decision to restrict access to its Mythos cybersecurity AI, OpenAI has stepped in to provide UK financial institutions with access to GPT-5.5 Cyber, a specialized model for cyber defense. The move highlights growing competition among AI providers to serve the cybersecurity needs of the financial sector.

OpenAI launches GPT-5.5 Spud with agentic capabilities and higher pricing

OpenAI has released GPT-5.5 Spud, a new model variant that introduces agentic features and comes with a higher per-token price point. This launch signals OpenAI's push to monetize more autonomous AI capabilities.

OpenAI announces plans for AI-powered robots to rival Tesla Optimus

OpenAI CEO Sam Altman revealed that the company is developing AI-powered robots targeting both infrastructure and home tasks, positioning them as competitors to Tesla's Optimus. The move marks OpenAI's expansion from software into physical robotics.

Google eases Gemini usage limits after user complaints

Google has adjusted compute-based quotas for its Gemini AI services and introduced free Flash-Lite prompts to address user feedback regarding excessive rate limiting. The move aims to improve the developer and user experience for complex AI queries.

CMA-CGM deploys Mistral AI to 80,000 employees for business support

French shipping giant CMA-CGM is rolling out a new AI platform called MAIA, powered by Mistral, to 80,000 employees worldwide. The system is designed to assist with a wide range of business operations, marking a large-scale deployment of sovereign AI in the logistics sector.

AI developer tooling & infrastructure

DeepSeek++ 0.5.1 brings MCP protocol and agent capabilities to the web version

The latest release of DeepSeek++ (v0.5.1) introduces support for the Model Context Protocol (MCP) and agent-like functionalities within its browser-based interface, enabling autonomous tool calls and a coding experience comparable to Claude Code.

Cursor launches Jira integration for ticket handling

Cursor has released a new integration that allows its AI agent to directly take ownership of Jira tickets, reducing context switching for developers and streamlining issue resolution.

SkipLabs' Skipper coding agent ships complete backend services without developer review

A new AI coding agent called Skipper can generate a fully running backend service from a plain-language prompt without any human feedback or review, representing a closed-loop autonomous software development approach.

Memory OS: a 6-layer open-source memory stack for Hermes Agent

A new open-source project called Memory OS introduces a six-layer memory architecture for AI agents, including auto-curated LLM wikis and multi-layer recall, built on top of the Hermes Agent framework and designed for token efficiency.

Open Envelope schema aims to be the 'Dockerfile' for AI agent teams

Open Envelope proposes a standardized JSON schema (team-v1.json) for defining AI agent teams with enforced access policies, controlling which components can call which APIs. The goal is to simplify the configuration and security of multi-agent systems.

Itential FlowAI general availability brings governed AI agents to enterprise infrastructure

Itential has announced the general availability of FlowAI, a platform that enables enterprises to design, deploy, and run AI agents on their infrastructure with built-in governance and security controls.

Cloud & platform providers

No significant new developments.

AI policy, regulation & governance

Anthropic gives EU cybersecurity agency ENISA access to Mythos AI via Project Glasswing

The European Union Agency for Cybersecurity (ENISA) has become the first EU institution to gain controlled access to Anthropic's advanced Mythos AI system under the Project Glasswing initiative. The access is intended to allow ENISA to test and evaluate Mythos for cybersecurity purposes before any broader release, marking a significant step in governance and regulatory engagement for frontier AI models.

Florida sues OpenAI and Sam Altman over chatbot safety concerns

The state of Florida has filed a civil lawsuit against OpenAI and CEO Sam Altman, alleging that the company prioritized profit over safety and violated consumer protection laws. The suit seeks penalties and aims to impose data collection limits for users under 13.

Google DeepMind's hurricane forecast model lauded by National Hurricane Center

The National Hurricane Center has publicly praised Google DeepMind's AI hurricane forecasting model, noting that it improved both lead time and confidence during the 2025 season. The endorsement signals a growing acceptance of AI in critical weather prediction.

US Treasury closes consultation on GENIUS Act principles for state AI regimes

The U.S. Department of the Treasury has concluded its public consultation on principles for determining whether state-level AI regulatory regimes are similar to the federal framework under the GENIUS Act. This marks a key step in aligning state and federal AI governance.

India to retain AI labelling rule and introduce schedule for advisories

The Indian government plans to keep a mandatory AI content labelling requirement as part of upcoming amendments to its IT rules, which are now out for public consultation. The move aims to increase transparency around AI-generated media.

South Korea privacy regulator reviews Naver AI Tab search agent

The South Korean privacy watchdog is conducting a prior adequacy review of Naver's AI Tab, a personalized AI search agent. The review aims to establish privacy conditions before the service is publicly launched.

Australian agencies falsely accused of flouting AI rules due to administrative errors

Several Australian government agencies, including the ABS and ACMA, were incorrectly flagged for non-compliance with AI transparency obligations, leading to public record corrections. The errors were caused by administrative mistakes, not actual violations.

Industry & market moves

Google forms special coding team after Mythos AI alarm, led by Sergey Brin

In response to the perceived threat from Anthropic's Mythos AI, Google co-founder Sergey Brin has reportedly established a dedicated 'coding special forces' team and a DeepMind sprint group to accelerate AI coding capabilities. The move underscores the competitive urgency triggered by breakthrough AI systems.

Revealed: internal power struggle at OpenAI that led to leadership change

A detailed report has emerged shedding light on a 72-hour internal conflict at OpenAI that resulted in a sudden leadership change. The account reveals deep team divisions and structural shifts within the company.

Salesforce acquires Contentful to bolster Agentforce AI platform

Salesforce has acquired Berlin-based headless CMS provider Contentful. The deal will add Contentful's content layer to Salesforce's Agentforce, aiming to enhance AI-driven enterprise content delivery.

Foxconn partners with France's Bull for AI expansion in Europe

Hon Hai Precision Industry (Foxconn) and French IT firm Bull have formed a partnership to build AI infrastructure across Europe, with initial manufacturing and testing facilities in the Czech Republic and France, totaling over €120 million in investment.

Mecka AI raises $60M Series A to train robots using human data from iPhones

Mecka AI has secured $60 million in Series A funding to develop a data platform that trains robots using human movement data collected from sensors and iPhones. The investment reflects growing interest in data infrastructure for robotics AI.

Geordie AI raises $30M Series A to secure enterprise AI agents

Geordie AI has raised $30 million in Series A funding to expand its platform that provides security and governance tools for enterprise AI agents. The funding will be used to scale its product and go-to-market efforts.

SOLAI to acquire 51% stake in NEURALAND, pivoting to AI-first platform

SOLAI Limited has announced an agreement to acquire a majority stake in NEURALAND, shifting its business from digital asset mining to an AI-first platform focused on hardware and software capabilities.

Robo.ai completes 100% acquisition of Neurovia AI, strengthening AI software foundation

Robo.ai Inc. has finalized its full acquisition of Neurovia AI Limited, solidifying its transition into an AI infrastructure platform. The deal aims to enhance Robo.ai's AI software capabilities.

Datavault AI signs $2 billion structured financing term sheet for digital assets

Datavault AI Inc. has signed a $2 billion structured financing term sheet, intended to create a new collateral base for its digital asset initiatives. The deal includes an exclusive global tokenization mandate for the counterparty.

Ant Group and Didi invest in embodied AI data company Jianzhi Robotics

Jianzhi Robotics, a company focused on 'body-agnostic data' for embodied AI, has secured its largest funding round to date with investments from Ant Group and Didi. The funding will support global expansion and data infrastructure development for embodied models.

AI product & feature launches

Workday and Google Cloud deepen AI agent partnership

Workday and Google Cloud have expanded their partnership to integrate Gemini Enterprise with Workday's Sana Self-Service Agent, enabling more seamless AI-driven HR and finance workflows directly within widely used Google Cloud tools.

Royal MCP plugin connects WordPress securely to AI models

A new WordPress plugin called Royal MCP provides an authenticated, rate-limited Model Context Protocol server, allowing AI agents like Claude, ChatGPT, and Gemini to safely interact with WordPress sites for data retrieval and management.

Enterprise software leaders adopt NVIDIA Agent Toolkit for AI development

Several major enterprise software companies have partnered with NVIDIA to leverage its Agent Toolkit, which includes the Nemotron models and NemoClaw ecosystem, to build and deploy autonomous AI agents.

Nvidia officially launches RTX Spark chip for Windows-based AI laptops

Nvidia has introduced the RTX Spark chip, designed to bring advanced agentic AI capabilities to Windows laptops. The chip is built on a 3nm process in partnership with MediaTek and targets the mainstream PC market.

NVIDIA launches Vera CPU, first purpose-built CPU for autonomous AI agents

NVIDIA has announced the Vera CPU, now in full production, claiming it is the first processor designed specifically for agentic AI workloads. It promises up to 1.8x faster task completion for autonomous AI tasks.

Autobrains and Uber to launch agentic AI robotaxi program in Munich on NVIDIA DRIVE

Autobrains and Uber are collaborating to deploy a jointly developed agentic AI robotaxi program in Munich, Germany, using NVIDIA DRIVE Hyperion platform. This marks the first European deployment of an autonomous driving system powered by advanced AI agents.

NVIDIA launches Factory Operations Blueprint for AI-powered manufacturing

NVIDIA has unveiled a new reference blueprint that gives factories an on-device AI brain to optimize operations, leveraging its hardware and AI models to improve efficiency and automation in manufacturing settings.

Nvidia and Unitree release H2 Plus open humanoid robot reference design

Nvidia and Unitree have jointly released the H2 Plus, an open reference design for humanoid robots built on the NVIDIA Isaac GR00T platform. The collaboration aims to accelerate research and development by unifying Unitree hardware and Nvidia software.

NVIDIA Alpamayo 2 Super open reasoning model targets safer robotaxis

NVIDIA has released Alpamayo 2 Super, a 32B-parameter open reasoning model designed to enhance the safety and scalability of level 4 robotaxis by providing full driving-stack reasoning capabilities.

Nvidia ramps production of Vera Rubin platform for next-gen AI factories

Nvidia has begun volume production of its Vera Rubin platform, which is designed to serve as the computational foundation for enterprise AI factories and autonomous AI agent workloads.

Nvidia DGX Station for Windows delivers 1 trillion-parameter AI in deskside form

Nvidia has introduced the DGX Station for Windows, a deskside AI supercomputer capable of running 1 trillion-parameter models, developed in collaboration with Microsoft. This brings massive AI computing power to individual workstations.

Zoom launches ZoomMate, an AI teammate that turns conversations into completed work

Zoom has introduced ZoomMate, described as the first AI teammate designed to transform workplace conversations into actionable tasks by integrating with multiple enterprise tools. The platform aims to automate meeting follow-ups and workflow execution.

Research with immediate practical relevance

Claude Opus 4.8 tops intelligence index while Mythos dominates hacking

According to a new benchmark report, Claude Opus 4.8 has achieved the top score on a general intelligence index, while Anthropic's Mythos system remains the leading tool for cybersecurity-related tasks. The report underscores the dual capabilities of Anthropic's models across both general and specialized domains.

Research finds frontier AI models fail EU law in up to 93% of tests

A new study has found that all tested frontier AI models violate EU legal requirements in up to 93% of scenarios, particularly around transparency and explainability. The findings highlight a significant compliance gap for AI developers and deployers operating in the EU.