AI news5
Google is investing up to $40B in Anthropic across a multi-year commitment
Google announced on Friday 24 April that it will invest $10B in Anthropic now in cash at a $350B valuation, with up to $30B more if performance targets are hit. The deal sits on top of Anthropic's existing multi-gigawatt TPU commitment starting 2027 through the Broadcom-Alphabet joint venture. It lands the same week Anthropic was reported pushing toward an IPO valuation near $800B with revenue tripling to $30B this year.
9to5google.com ↗Why it matters
The frontier-lab capital structure is consolidating around hyperscaler strategic investment, which compresses exit optionality for labs and raises the stakes on sovereign and regional AI capacity.
Anthropic turned Claude into a consumer front door with 15 app connectors
Anthropic announced on Friday 24 April a wave of consumer app connectors for Claude covering Spotify, Uber, Uber Eats, Instacart, Intuit TurboTax, Intuit Credit Karma, TripAdvisor, Booking.com, AllTrails, Audible, Resy, StubHub, Taskrabbit, Thumbtack, and Viator. Claude surfaces the right app in-conversation based on what the user is doing, with user data from connected apps not used to train models and no sponsored placements. The company restated that Claude will stay ad-free.
pymnts.com ↗Why it matters
This is Anthropic's most explicit consumer move to date and positions Claude as an assistant layer across daily services, breaking out of developer/enterprise-only positioning.
DeepSeek previewed V4 Flash and V4 Pro with a 1M-token context window
DeepSeek unveiled preview versions of its V4 Flash and V4 Pro flagship models on Friday 24 April, claiming top-tier coding-benchmark performance and major gains on reasoning and agentic tasks. The release leans on a new Hybrid Attention Architecture for cross-turn memory and pushes the context window to 1 million tokens. DeepSeek is framing V4 as the most powerful open-source AI platform available, running on Huawei chips with "full support" from the vendor.
fortune.com ↗Why it matters
An open-source frontier-class model running on non-Nvidia silicon narrows both the cost and the geopolitical gap in the frontier-model stack, and keeps downward price pressure on closed US-lab pricing.
Anthropic denied intentional Claude Code slowdown and shipped three fixes
Anthropic publicly addressed user complaints on Friday 24 April that Claude Code had been quietly degraded, confirming it identified and patched three performance issues. The company denied any intentional "nerfing" and reversed the Pro-plan Claude Code pricing test from earlier in the week. User complaints had been building on X and Hacker News since the Opus 4.7 GA release.
dataconomy.com ↗Why it matters
Reputational quality incidents at the frontier labs are now direct community-pressure events rather than silent engineering work, and how labs communicate through them is becoming part of the product.
Google turned Chrome into an enterprise AI coworker with Gemini auto-browse
Google highlighted this week that Gemini Enterprise is now processing more than 16 billion tokens per minute via direct API usage, and rolled out Gemini auto-browse in Chrome for enterprise users. The feature lets Gemini understand what's happening across open tabs and act on tasks like booking travel, comparing vendors, and scheduling meetings.
promethean-ai.com ↗Why it matters
Browser-resident AI agents with multi-tab awareness is the shortest path from current chat-based copilots to true workflow automation, and Chrome is the default surface for most enterprise knowledge workers.
AI in the nonprofit sector4
Google.org AI for Science Impact Challenge deadline is 1 May, $500K–$3M per award
The Google.org Impact Challenge: AI for Science closes applications on Thursday 1 May 2026 with a $30M global pool. Grants range from $500K to $3M each, open to nonprofits, social enterprises, and academic institutions. Selected organisations also receive a place in the Google.org Accelerator, which provides technical expertise on generative and agentic AI capabilities.
google.org ↗Why it matters
This is the highest per-award ceiling currently open for nonprofit AI funding, and the accelerator component closes the "we have the grant but not the technical capacity" gap that typically sinks first-year AI implementation.
Microsoft Elevate for Changemakers launched free AI-readiness credentials for nonprofit leaders
Microsoft Elevate for Changemakers opened in April 2026 to support nonprofit and social-impact leaders driving practical AI adoption. The programme builds hands-on skills through free AI-readiness credentials, focused on strengthening internal capability so AI can be used safely, effectively, and in line with community expectations. It sits alongside Microsoft's broader nonprofit cloud and AI credit programmes, and is delivered through Microsoft Elevate led by President Justin Spelhaug.
news.microsoft.com ↗Why it matters
Free, credential-bearing training at the leadership tier lands exactly where the Virtuous/Fundraising.AI adoption report said orgs were stuck — 81% of nonprofits use AI individually without shared workflows, which leader-level competence is needed to fix.
Fund.AI second-annual convened 200 philanthropic leaders on AI for good
The Patrick J. McGovern Foundation and Fast Forward hosted the second-annual Fund.AI convening this month, a two-day in-person gathering of 200 philanthropic leaders on funding nonprofit AI responsibly. Programme included sessions with OpenAI, Anthropic, Google.org, and practitioners on how funders can support nonprofits building and using AI. McGovern Foundation has committed $75.8M across 149 grants in 2025 alone for AI-for-humanity work.
candid.org ↗Why it matters
The funder-side conversation is now consolidated enough to convene at scale, and the post-convening output usually shapes what frames show up in RFPs over the following six months.
Tech nonprofit coalition pushed back on US GSA AI procurement rules
The Electronic Frontier Foundation and allied tech nonprofits responded publicly this month to proposed General Services Administration rules intended to steer federal AI spending toward "ideologically neutral" American AI innovation. The coalition argues the rules go beyond their stated goal and could be used to undermine the AI trust and safety standards that public-interest tech organisations advocate for. The statement lands parallel to California's Executive Order N-5-26 imposing new trust-and-safety obligations on AI vendors.
eff.org ↗Why it matters
Federal procurement is the single biggest lever for shaping which AI vendors nonprofits, universities, and state agencies can actually buy from, so the terms being set now will define the next five years of sector-accessible AI.