The Informed Layer · Newsscan
Daily briefing

Anthropic expands in APAC; EU weighs safety carve-out

← All issues

Anthropic / Claude ecosystem

Anthropic signs its first Australian data centre agreement

Anthropic has signed its first data centre agreement in Australia, taking capacity in a large planned renewable-powered campus near Brisbane. The deal remains subject to Foreign Investment Review Board approval.

Anthropic to open Singapore office, its fifth in Asia-Pacific

Anthropic will open a Singapore office in October, its fifth Asia-Pacific location, citing Singapore's among-the-highest per-capita Claude usage worldwide as a standout market. The move deepens regional enterprise support.

Anthropic and Novo Nordisk announce AI drug discovery collaboration

Novo Nordisk, maker of Ozempic, will partner with Anthropic to apply Claude models to drug discovery, with the stated ambition of becoming the world's most AI-driven healthcare company.

Anthropic adds new proactive safeguard to Claude Fable 5

Anthropic has introduced a new proactive guardrail on Claude Fable 5 that restricts access to restricted capabilities before misuse occurs. The move is being read as part of Anthropic's positioning with the Commerce Department amid ongoing regulatory interactions.

Frontier model providers

OpenAI launches ChatGPT Ads with Sponsored Agents and AI-assisted ad tools

OpenAI unveiled its advertising platform for ChatGPT, including Sponsored Agents, AI-assisted ad creation and optimisation, Ads Manager enhancements, and integrations with HubSpot and Shopify.

Google and DeepMind launch institute to explore AGI

Google and DeepMind are launching a cross-organisational institute to explore the implications of artificial general intelligence, with research to be made publicly accessible. Co-founder Shane Legg is closely associated with its safety-first framing.

AI developer tooling & infrastructure

Zed launches Delta in public beta to replace pull requests

Zed's Delta, built on DeltaDB, enters public beta replacing pull requests with thread-based collaboration over shared worktrees and agent-assisted reviews, modelling software state as deltas rather than commits.

ONES MCP connector lands in Claude's connectors directory

ONES released an MCP connector now listed in the Claude Connectors Directory, giving Claude direct access to ONES project-management data and reducing time lost switching between tools.

TermSquad launches an always-on cloud computer for AI coding agents

TermSquad introduced a persistent cloud Linux environment purpose-built for AI coding agents, offering always-on sessions and multi-agent orchestration for long-running development workflows.

Meta launches MCP server for WhatsApp Business setup and management

Meta introduced a Model Context Protocol server for the WhatsApp Business Platform, letting AI agents configure accounts, templates and webhooks, and test setups without switching between consoles.

WSO2 launches Agent Manager for sovereign governance of enterprise AI agent sprawl

WSO2 introduced Agent Manager, an open-source, model- and framework-agnostic platform to govern large, heterogeneous AI agent estates across environments, pitched as sovereign AI governance for enterprises.

Cloud & platform providers

Vercel AI SDK Harness adds native subscription sign-in for coding agents

The Vercel AI SDK harness layer now supports host-based native subscription authentication, letting coding agents onboard through existing subscriptions such as Claude Code without code changes.

AI policy, regulation & governance

EU open to antitrust exemption for AI safety coordination, Ribera says

The EU would consider relaxing antitrust rules to allow AI firms to coordinate on safety if a formal request is made, according to Executive Vice-President Teresa Ribera. This would clear a legal obstacle for the joint industry safety body the major labs have discussed.

Von der Leyen to invite AI leaders to support efforts to 'pace the frontier'

European Commission President Ursula von der Leyen will invite leading AI developers to back international efforts to pace frontier model development. The initiative is tied to European priorities in health, transport, agri-food, manufacturing and defence/space.

UN chief sounds alarm on AI risk ahead of General Assembly

UN Secretary-General António Guterres urged global governance and stronger AI oversight ahead of the UN General Assembly, warning of rapidly advancing AI risks after President Trump played the dangers down.

'If you're building Frankenstein, stop': JD Vance dismisses calls for AI regulation

US Vice President JD Vance publicly dismissed calls for AI regulation, telling developers that if they are 'building Frankenstein' they should stop rather than seek new rules. His comments add White House weight against regulatory approaches in the escalating AI-safety debate.

First agentic AI data breach reported to Spanish regulator

The first known data breach executed by an autonomous AI agent has been reported to Spain's data protection authority. The incident marks a new class of AI-enabled attack and is expected to drive elevated, AI-assisted risk management and incident-response requirements.

Industry & market moves

Mozilla partners with Mistral to power Firefox's AI features

Mozilla announced a partnership with Mistral to power Firefox's AI assistant (Smart Window) with Mistral Small 4, framing the deal around user choice, openness and interoperability, with expanded French-language support and a broader rollout to come.

OpenText and Cohere partner to combine trusted data with agentic AI

OpenText and Cohere announced a partnership pairing OpenText's enterprise data and context layer with Cohere's agentic AI platform, described as first-of-its-kind for regulated sectors, with deployments starting early 2027.

ByteDance's AI drug unit Anew Labs completes $290M raise after spin-off

Anew Labs, ByteDance's AI drugmaking unit, completed a $290 million fundraising round at a $1.5 billion valuation in its first external raise following its spin-off, sources said.

Japan's Resonac and Nitto Denko to invest in US fund for AI hardware

Japanese chip-material makers Resonac Holdings and Nitto Denko will invest in a US-based fund targeting AI hardware production bottlenecks, deepening Japan-US supply chain collaboration.

Mitsubishi Heavy to deepen Preferred Networks AI tie-up with new investment

Mitsubishi Heavy Industries will invest to expand its collaboration with Preferred Networks, its first publicly reported step to scale the partnership, with AI defence technology among the focus areas.

LawZero secures up to $300M in joint Canada–Germany funding

Yoshua Bengio's AI safety nonprofit LawZero received a commitment of up to CAD $300 million in joint funding from Canada and Germany to advance its Scientist AI approach to safe, trustworthy next-generation AI and sovereign AI capacity.

Barndoor acquires Diaphora to bring governed AI automation to enterprise workflows

Barndoor acquired Diaphora to strengthen governed AI automation across enterprise tools and data, adding governed Blueprints and role-based access controls for reliably scalable workflow automation.

Former Infosys chief's AI startup Hang Ten adds $53M weeks after initial seed

Hang Ten, the enterprise AI software-development startup founded by the former Infosys chief, closed an additional $53 million weeks after its initial raise, bringing total seed funding to $85 million amid strong enterprise interest.

Arcee AI hits $1 billion valuation after training four models for $20 million

Arcee AI closed a Series B at a $1 billion pre-money valuation, having spent only $20 million to train four open-weight models with competitive benchmark results.

OpenAI reportedly eyes a $1.2 trillion funding round

OpenAI is reportedly discussing a funding round that could push its valuation above $1.2 trillion ahead of an anticipated IPO. The figure would dwarf any prior private raise.

Headspace being acquired by AI health startup Sword for $300 million

Sword Health, an AI-driven healthcare startup, is acquiring meditation app Headspace in an all-cash deal near $300 million, with closing expected by early Q4 2026.

AI product & feature launches

Simplilearn and Carnegie Mellon launch executive program in LLMs and multi-agent systems

Carnegie Mellon's School of Computer Science Executive Education and Simplilearn launched a joint program to formalise professional training in LLMs and multi-agent systems, aimed at moving organisations from experiments to reliable deployed agent stacks.

ShengShu launches Vidu S2 with real-time interaction and editing for AI video

ShengShu Technology unveiled Vidu S2, adding real-time interactive generation and editing via S2-Avatar and S2-Editing, with higher resolution, dynamic reference image updates and VR spatial-video integration.

NVIDIA's Vera Rubin NVL72 tops MLPerf Inference v6.1 in first showing

NVIDIA's Vera Rubin NVL72 delivered leading inference performance in its first MLPerf Inference v6.1 preview submission, outperforming the GB300 NVL72 generation.

Rubrik launches Code Guardian, red-teaming code with Claude Mythos 5

Rubrik unveiled Code Guardian, a frontier-AI powered harness that uses Anthropic's Claude Mythos 5 to red-team air-gapped repositories and prioritise exploitable vulnerability chains for remediation.

Research with immediate practical relevance

Nature: AI tool turns any research paper into a collaborating 'agent'

A tool described in Nature converts static papers into interactive 'paper agents' that answer complex queries, apply the paper's methods to new data, and collaborate with other paper agents to surface new discoveries, built on the Model Context Protocol.

IonQ, ORNL and NVIDIA show generative AI accelerates quantum optimisation

IonQ, Oak Ridge National Laboratory, NVIDIA and the University of Tennessee demonstrated that generative AI can directly produce quantum optimisation circuits, reducing iterative tuning costs and enabling better scaling of hybrid quantum workloads.

ZGCM-1 is a fully open foundation model for mathematics and agentic search, using mixed SFT/RL training with ACE diagnostics and gated sliding-window attention for efficiency, developed through an AI-native R&D workflow.

LLaDA-UI: a 16.7B diffusion-based GUI agent

LLaDA-UI is a 16.7B-parameter GUI agent that performs diffusion-based reasoning and outputs structured GUI actions, applying diffusion language modelling to computer-control tasks.