Anthropic expands; Regulators eye AI agents and risk

Anthropic / Claude ecosystem

Anthropic Shrugs Off $800B+ Valuation Offers for Now, Eyeing IPO in 2026

Anthropic is reportedly not currently interested in new funding offers that would value the company at $800 billion or more, despite a recent report of seeking such investment. The company has significant capital expenditures for data centers and cloud services, and is in early talks for a potential IPO as early as October 2026.

Anthropic Launches Claude for Word with Tracked Changes, Comment Replies, and Document Scanning

Anthropic has introduced 'Claude for Word', a Microsoft Word integration allowing users to leverage Claude's editing and content generation features directly within documents. It supports tracked changes, in-thread comment management, document scanning for issues, and reusable workflows, available in beta for Team and Enterprise plan subscribers.

Frontier model providers

Five9 Introduces Agent Assist 2.0 with AI Summary Powered by OpenAI

Five9 has released Agent Assist 2.0, featuring an AI Summary powered by OpenAI's generative AI technology. This solution summarizes customer call transcripts in seconds, eliminating the need for manual after-call work and model training, and is part of Five9's strategy to democratize advanced AI for contact centers.

AI developer tooling & infrastructure

Cursor 3.1 Released with Parallel Agents and Improved Voice Input

Cursor, an AI-first code editor, has released version 3.1, featuring an enhanced Agents Window with a 'tiled layout' for running multiple AI agents in parallel and comparing their outputs. The update also includes a rewritten voice input system for better accuracy and the ability to select specific Git branches for cloud agents.

Cloud & platform providers

AWS Interconnect Now Generally Available for Multi-Cloud and Last-Mile Connectivity

AWS has announced the general availability of AWS Interconnect, offering two new capabilities: Multi-Cloud, providing managed private connectivity between AWS VPCs and other cloud providers (starting with Google Cloud, Azure later in 2026); and Last Mile, simplifying high-speed, private connections from on-premises networks to AWS through existing network providers. Both are fully managed and secure.

Cloudflare Launches Mesh for Secure Private Networking Across Users, Nodes, and AI Agents

Cloudflare has introduced Cloudflare Mesh, a new private networking solution designed for the era of AI agents. It unifies AI agents, humans, and multi-cloud infrastructure into a single secure fabric, enabling agents to securely access private databases, internal APIs, and staging environments, integrated with Cloudflare's Developer Platform and One security controls.

Cloudflare Adds Managed OAuth for Access to Make Internal Apps Agent-Ready in One Click

Cloudflare has enabled managed OAuth for all Access applications, allowing AI agents that speak OAuth 2.0 to easily authenticate and access internal apps, web pages, and REST APIs. This eliminates the need for service accounts with static credentials, improving security and auditability.

AI policy, regulation & governance

Washington State Enacts New AI Laws Targeting Chatbots and Misinformation

Washington state has passed new legislation regulating AI, requiring greater transparency from companies like OpenAI and Anthropic regarding chatbot operations. One law mandates watermarks or metadata for significantly altered generative AI content, while the second requires chatbots to disclose their non-human nature and prohibits sexually explicit conversations with minors.

Italian President Mattarella Urges AI Regulation to Prevent Big Tech Domination

Italian President Sergio Mattarella has called for clear rules on artificial intelligence to prevent it from becoming a tool for domination by large technology companies. He emphasized the need for 'adequate moral awareness' and regulations, citing existing EU efforts, to govern AI and prevent it from undermining sovereign states and the international order.

South Africa Publishes Draft National Artificial Intelligence Policy for Public Comment

South Africa's Department of Communications and Digital Technologies has published its Draft National AI Policy for public comment, outlining AI as a tool for inclusive economic growth. The policy identifies priority areas like education and healthcare, proposes new institutional oversight mechanisms, and envisages a phased, risk-based approach to AI regulation.

Council of Europe Unveils New Platform Regulation Recommendations, Focusing on Human Rights

The Council of Europe has released new policy recommendations for platform regulation, emphasizing a human rights approach and reconceiving user empowerment as the starting point for regulation. The 'Recommendation on the online safety and empowerment of users and content creators' calls for middleware-based content curation, collective redress mechanisms, and child-centric age assurance, building on existing European frameworks.

Australian National Audit Office Scrutinizes DoHDA's AI Use for Health Provider Non-Compliance

The Australian National Audit Office (ANAO) has launched an audit into the Department of Health, Disability and Ageing's (DoHDA) use of AI to manage health provider non-compliance. The audit will assess DoHDA's AI governance, design, development, deployment, monitoring, and reporting, with contributions accepted until April 26, 2026.

Australian Financial Regulators in High-Level Talks with Banks and Super Funds Over Anthropic's Mythos Risks

Australia's major banks, insurers, and super funds are engaged in urgent discussions with financial regulators (APRA and ASIC) regarding the cybersecurity risks posed by Anthropic's powerful Mythos AI model. Despite Australia not receiving early access, regulators are pressing institutions to fortify defenses against potential weaponization of Mythos's vulnerability-finding capabilities.

Industry & market moves

Allbirds Pivots to AI Compute Infrastructure, Rebrands as NewBird AI with $50M Convertible Financing

Following the sale of its footwear brand and assets, Allbirds is pivoting its business to AI compute infrastructure, rebranding as 'NewBird AI.' The company secured a $50 million convertible financing facility to acquire high-performance GPU assets and aims to become a GPU-as-a-Service and AI-native cloud solutions provider.

OpenGradient Announces $9.5 Million in Funding to Build Compute Layer for Verifiable AI

OpenGradient has raised $9.5 million in total funding, backed by a16z crypto and other investors, to scale its 'Network for Open Intelligence.' This decentralized infrastructure aims to host, execute, and verify AI models at scale by attaching cryptographic proofs to every inference, enabling auditable model execution for developers.

University of Chicago Partners with Microsoft, NVIDIA, and AI Research Commons to Accelerate Midwest AI Startups

The University of Chicago, through its Polsky Center, has partnered with AI Research Commons (ARC), Microsoft, and NVIDIA to support early-stage AI startups from Third Coast Foundry universities. Selected startups will receive technical and go-to-market support, up to $350K in Azure credits, NVIDIA Inception program benefits, and connections to Bay Area investors.

Obriy AI Raises $500K to Take Multi-Agent Enterprise Automation Global

Ukrainian startup Obriy AI has raised $500,000 from N1 Investment Company to expand its multi-agent AI platform, SURE, for enterprise automation. SURE orchestrates LLMs and data to automate complex business processes, from customer support to legal queries, and has been piloted with Ukraine's Ministry of Justice.

HubSpot Launches AEO (Answer Engine Optimization), AI Agents, and Smart Deal Progression

HubSpot has introduced HubSpot AEO (Answer Engine Optimization) to help marketers optimize their visibility in answer engines like ChatGPT and Gemini. The company also expanded its Prospecting Agent for full lifecycle management and updated its Customer Agent for email support, alongside Smart Deal Progression for sales.

Oracle Financial Services Extends Agentic AI Platform to Corporate Banking

Oracle Financial Services has launched new embedded AI capabilities and pre-built AI agents for corporate banking, including treasury, trade finance, credit, and lending. These advancements automate mission-critical processes like loan data extraction and validation, and streamline decision-making with human oversight.

Research with immediate practical relevance

NVIDIA and University of Maryland Release Audio Flamingo Next (AF-Next), a Powerful Open Large Audio-Language Model

NVIDIA and the University of Maryland have released Audio Flamingo Next (AF-Next), a new open Large Audio-Language Model (LALM) trained on internet-scale audio data. AF-Next comes in three specialized variants (Instruct, Think, Captioner) and features Temporal Audio Chain-of-Thought for robust reasoning over long audio, outperforming closed models like Gemini 2.5 Pro on long audio benchmarks.

Lantern Pharma Launches withZeta.ai, a Multi-Agentic AI Co-Scientist for Rare Cancer Drug Discovery

Lantern Pharma has commercially launched withZeta.ai, described as the world's first multi-agentic AI co-scientist for rare cancer drug discovery and development. The platform offers subscription tiers and utilizes a multi-agentic, multi-tool architecture grounded in Lantern Pharma's proprietary rare cancer ontology to query databases, scientific literature, and molecular data, providing actionable insights.