AI news5
DeepSeek closes $7.4B in first-ever external funding at a $50B+ valuation — with founder staying firmly in control
Chinese AI lab DeepSeek has raised more than 50 billion yuan ($7.4 billion) in its first external funding round, with a valuation landing between $52 billion and $59 billion according to Reuters. The deal structure is unusual: investors put capital into a limited partnership managed by CEO Liang Wenfeng rather than into DeepSeek itself, with no voting rights and a five-year lock-up. Liang committed 20 billion yuan of his own money — the controlling share of the raise — with Tencent and battery maker CATL among the largest outside backers. China's state AI investment fund is the only investor with voting rights and no lock-up. DeepSeek had previously been funded entirely from the balance sheet of Liang's quant hedge fund High-Flyer, and this round is less an opening than a carefully managed crack.
finance.yahoo.com ↗Why it matters
a $50B+ valuation for a lab that has never taken VC before — and whose V4 model runs on Huawei chips at a fraction of US frontier costs — signals that China's AI industry is entering a new capital formation phase, with deal structures designed to preserve founder and state control simultaneously.
Microsoft Copilot Cowork goes globally available for enterprise customers with Anthropic Claude integration
Microsoft made Copilot Cowork generally available worldwide for Microsoft 365 customers on 16 June after a three-month Frontier preview. More than half of the Fortune 500 used Cowork during the preview, including Accenture, Capital Group, Koch, and Zurich Insurance. Cowork executes complex, long-running tasks across tools and business systems — editing batch job spreadsheets, generating dependency flow charts, comparing thousands of files, and reviewing stalled sales pipelines — rather than returning drafts for humans to act on. The system runs inside Microsoft 365, uses Work IQ to ground tasks in company data, and all prompts and artifacts are governed by existing Microsoft 365 audit, eDiscovery, and data security controls.
cryptobriefing.com ↗Why it matters
Fortune 500 adoption during preview plus global GA marks the moment Microsoft 365 shifts from AI-assisted drafting to AI-executed task completion at scale — and the Anthropic Claude integration broadens the model options enterprise customers can ground against their own data.
Databricks launches Genie One, Genie Agents, and Genie Ontology as a unified AI data coworker for business
Databricks launched three interconnected products on 16 June: Genie One, an AI coworker for business users that goes beyond conversational analytics to take agentic actions across Slack, Teams, mobile, and MCP-based experiences; Genie Agents, which lets teams turn prompts into shareable autonomous agents that can reason over structured and unstructured data; and Genie Ontology, a unified context layer that extracts knowledge from data and connected tools so agents can answer with full business context. The launch addresses a persistent gap: current agents often iterate slowly when business context is scattered across dashboards, wikis, tickets, and chat threads, producing generic or wrong answers.
databricks.com ↗Why it matters
Databricks is repositioning the entire data platform layer as the ground-truth foundation for agentic AI — effectively arguing that trustworthy enterprise agents require an ontology tied to the business's actual data, not just a general-purpose LLM surfaced via a chat interface.
Google DeepMind and partners open a $10M research call to study what happens when millions of AI agents interact
Google DeepMind, Schmidt Sciences, the Cooperative AI Foundation, ARIA, and Google.org launched a $10 million funding call for technical research into multi-agent AI safety. The call targets what happens when large numbers of AI agents, built and deployed by different organizations, interact across shared digital environments — risks that emerge from agent populations rather than from individual models in isolation. Proposals are due 8 August 2026, with awardees notified in fall 2026. The initiative accompanies a DeepMind paper warning that a "multi-agent internet" — millions of agents operating autonomously at speed and scale — may emerge faster than safety researchers are prepared for.
edtechinnovationhub.com ↗Why it matters
this is the first major coordinated research fund targeting the safety risks of AI agent ecosystems rather than individual models — a category of risk that becomes the dominant concern once enterprise and consumer agent deployments start interacting with each other rather than just with humans.
AlphaGo architect David Silver's Ineffable Intelligence names Google Cloud as its frontier infrastructure partner
Ineffable Intelligence, the London startup founded by DeepMind researcher and AlphaGo lead David Silver, has named Google Cloud as its infrastructure partner for its frontier AI lab — announced at Google Cloud Next on 16 June. Silver founded Ineffable in 2026 after raising a $1.1 billion seed at a $5.1 billion valuation, with the thesis that reinforcement-learning-only systems will overtake models trained on human data. Google Cloud was selected after a competitive infrastructure evaluation. The partnership gives Ineffable access to Google's TPU clusters and AI infrastructure at a scale suited to the compute-intensive RL training regime Silver is pursuing.
thenextweb.com ↗Why it matters
the world's most experienced reinforcement-learning researcher has now publicly committed to a compute stack and lab structure for what he believes is the path to AGI — a concrete research bet that the post-human-data training era Silver describes is now being capitalised and resourced seriously.
AI in the nonprofit sector4
Nature Health study: Ada's clinical AI on South Africa's MomConnect prompted treatment for half of pregnant women who had not planned to seek care
A peer-reviewed study published in Nature Health on 16 June found that more than half of pregnant women and mothers in South Africa who received treatment after using Ada Health's symptom assessment tool on the government's MomConnect platform had not planned to seek care before using the AI. The SAFEMOM study — the largest clinical AI evaluation on MomConnect to date — found Ada improved safe and appropriate health-seeking behaviours, with the greatest impact among women facing the most barriers to care: lower health literacy, limited facility access, and first pregnancies. MomConnect reaches millions of pregnant women via basic phone channels across South Africa's public health system.
prnewswire.com ↗Why it matters
a Nature Health-published RCT showing that a clinical AI symptom tool on a government mass-health platform prompted over half its treated users to seek care they otherwise would have skipped is a direct answer to the question of whether AI in low-resource public health infrastructure changes outcomes — and it does, particularly for the highest-need users.
India's one million ASHA workers are co-designing the AI pregnancy tracker meant to support them
Researchers from the George Institute for Global Health and the University of Oxford are building SMARThealth Pregnancy 2, an AI-powered app to help ASHA workers identify and manage high-risk pregnancies in rural India — but unlike most digital health tools, ASHA workers are co-designers from the start. The co-design process involves field observation, iterative interface prototypes, and structured feedback from the workers themselves, who are shaping language, data presentation, and workflow to match the realities of home visits in communities with poor connectivity and high antenatal risk. The app will screen for hypertension, anaemia, and gestational diabetes in real time and flag referral cases. One million ASHA workers care for an estimated 25 million pregnant women annually across India's rural communities.
thebetterindia.com ↗Why it matters
a co-designed AI tool built by the people who will use it rather than handed down from a research institution is more likely to be adopted consistently — and in a system where ASHA workers are already the primary healthcare infrastructure for rural India, adoption is the difference between a tool that saves lives and one that collects dust alongside the paper registers it was meant to replace.
Carnegie Mellon and Meta partner to build AI tools that give emergency managers real-time population movement maps during disasters
Carnegie Mellon's NSF AI Institute for Societal Decision Making and Meta's AI for Good program announced a partnership on 16 June to develop dynamic situation reports for first responders during natural disasters including wildfires, hurricanes, and severe winter storms. NSF AI-SDM researchers will analyse aggregated mobility and connectivity data already collected through Meta's AI for Good program and turn it into clear visualisations that emergency managers can use on the ground during a disaster to track population movement and make faster decisions. The research team will work to make the tools accessible to local emergency agencies, not just federal or large-state operations.
cmu.edu ↗Why it matters
first responders in disasters routinely make evacuation and resource deployment decisions with hours-old or incomplete population data — AI-synthesised real-time mobility maps drawn from existing connectivity infrastructure could close that gap without requiring new sensor deployment in already-stretched emergency situations.
UAE convenes global AI dialogue on anticipatory humanitarian action — using AI to get ahead of crises before they hit
The UAE is hosting a high-level global dialogue on anticipatory humanitarian action at the Zayed National Museum in Abu Dhabi, announced 15 June. The summit brings together governments, UN agencies, and humanitarian organisations to explore how AI and predictive analytics can be deployed before crises escalate — moving from reactive disaster response to pre-positioned aid and early intervention. The dialogue focuses on how AI can translate early warning signals into faster decision cycles for humanitarian actors before displacement, food insecurity, or health crises become acute.
gulfnews.com ↗Why it matters
most humanitarian AI investment still focuses on response — what happens after a crisis hits. A senior dialogue convened around anticipatory action reflects a shift toward AI systems that can trigger pre-positioned resources, reducing both human cost and the scale of response needed once a crisis arrives.