AI news5
Canva launches AI 2.0 — "an AI platform with design tools" — with four new capabilities and six workflows
On April 16 at Canva Create in Los Angeles, Canva unveiled AI 2.0 as a full platform reset. CPO Cameron Adams framed it as "Canva reimagined as a conversational, agentic platform," with more than two years spent making Canva's design foundations accessible to AI. The assistant runs on the Canva Design Model — the first foundation model built specifically for real-world design — and ships with conversational design, agentic orchestration, layered object intelligence, and a persistent Memory Library, plus six new workflows including connectors, scheduling, web research, Brand Intelligence, Sheets AI, and Canva Code 2.0.
canva.com ↗Why it matters
Australia's biggest software company reframed from design tool to AI platform in a single event, and the Design Model is the first domain-specific foundation model from a non-US incumbent.
Anthropic crosses $30B run-rate, passing OpenAI's $25B for the first time
Anthropic's annualised run-rate revenue crossed $30B in April, according to TechCrunch and Bloomberg — putting it ahead of OpenAI's reported $25B ARR for the first time. Claude Code alone hit $1B ARR by November 2025 and more than doubled to $2.5B by February 2026. Anthropic is running at 10x annual growth for three consecutive years, the fastest B2B software scale trajectory on record.
the-ai-corner.com ↗Why it matters
Not vendor-hype — for practitioners building on Claude (Claude Code, Managed Agents, Bedrock), this is the signal that Anthropic's pricing, roadmap, and product-support commitments are now backed by the balance sheet to keep them stable.
OpenAI ships GPT-Rosalind, its first purpose-built domain model, as a restricted preview for life sciences
Released April 16, GPT-Rosalind is OpenAI's first purpose-built domain-specific model — engineered around biochemistry, genomics, protein engineering, and translational medicine workflows. Named for Rosalind Franklin, it is strongest on reasoning over molecules, proteins, genes, pathways, and disease-relevant biology, and on using scientific tools and databases in multi-step workflows. Launch partners are Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific; access is gated to qualified US enterprise customers doing legitimate life sciences research.
openai.com ↗Why it matters
First frontier-lab domain-specialised flagship — the pattern started with GPT-5.4-Cyber and now extends to life sciences, signalling that vertical domain models are the next product axis for OpenAI and Anthropic.
Google brings Gemini 3.1 Flash-Lite to general availability at $0.25 per million input tokens
Google's most cost-efficient frontier-tier model — Gemini 3.1 Flash-Lite — is now available through the Gemini API and Vertex AI at $0.25 per million input tokens and $1.50 per million output tokens. It delivers 2.5x faster Time to First Answer Token and 45% faster output generation than Gemini 2.5 Flash while matching or exceeding quality. Google is positioning it as the budget tier against OpenAI's GPT-5.4 mini and Anthropic's Haiku 4.5.
blog.google ↗Why it matters
Sets a new practitioner price floor for production-quality multimodal inference, which compresses the economics of high-volume agentic workloads like voice, translation, and bulk transcription.
Sona raises $45M Series B to bring AI to the "frontline economy" — scheduling, HR, payroll for real-economy businesses
Sona announced a $45M Series B this week, pitched as the operational foundation for frontline economy businesses — retail, hospitality, logistics, healthcare — where AI has historically been an awkward fit. Sona already powers scheduling, HR, and payroll for these customers and is compressing a ten-year product roadmap into twelve months. The round positions the company as a counterpoint to enterprise AI tooling that assumes desk-based knowledge workers.
prnewswire.com ↗Why it matters
First major VC bet on AI for shift-based and hourly-worker operations, a segment covering over 2 billion workers globally and largely untouched by the current agent wave.
AI in the nonprofit sector4
KPMG Foundation expands $6M AI Impact Initiative — 12 nonprofits backed, Salesforce and Braven join as partners
The KPMG US Foundation's AI Impact Initiative has now committed $6M across 12 US nonprofits to integrate AI into operations, paired with skills-based volunteering and pro-bono consulting. Recent grants include $500,000 each to First Book and Big Brothers Big Sisters, and $1.1M to Women's Health Access Matters for AI-powered health research. Salesforce and Braven have joined as collaboration partners, and free AI learning sessions have reached hundreds of nonprofit participants.
kpmg.com ↗Why it matters
One of the few corporate-foundation AI programs pairing capital with delivery support rather than just unrestricted grants, and the named nonprofit recipients are concrete enough to replicate the pitch template.
Opportunity International runs a local-language farmer chatbot in live deployment to scale agricultural guidance
Opportunity International has a local-language AI chatbot in active deployment giving smallholder farmers instant agricultural guidance — tailored to literacy levels and local languages — so frontline staff can focus on mentoring and long-term client relationships rather than repeat-FAQ work. The Chronicle of Philanthropy profiled the deployment alongside frontline AI work at Stop Soldier Suicide, IRC, and Samaritas. Funded through a mix of foundation grants and AWS credits.
mcgovern.org ↗Why it matters
Clearest live example to date of AI being used to expand reach for a global NGO without replacing the human relationship at the centre of the service model.
HEAL Capacity Grant closes April 30 — up to $20K for frontline community-based organisations
The Frontline Funding Collaborative's HEAL Capacity Grant (Healthy Environment for All) is accepting applications for up to $20,000 per grantee through April 30, 2026 — ten days out. The fund backs community-based organisations doing environmental-justice and health-equity work, with an explicit openness to AI and technology capacity as eligible expenditure. Prior rounds have funded environmental justice coalitions and frontline worker programs across multiple US states.
grantedai.com ↗Why it matters
Small but fast-cycle unrestricted capacity funding explicitly open to AI investment at the frontline level — the tier most nonprofit AI programs ignore.
Tech nonprofits warn federal agencies not to weaponise AI procurement against trust and safety
The Electronic Frontier Foundation published a position piece this week warning that US federal agencies are quietly rewriting procurement rules in ways that could punish AI companies for declining to allow their technology to be used for mass surveillance. The piece argues that procurement should pull in favour of open-source, interoperability, and right-to-repair — not against responsible-use positions taken by model providers. Follows the public dispute between the Department of Defense and Anthropic over surveillance use-cases.
eff.org ↗Why it matters
Sector-adjacent but directly relevant to any nonprofit negotiating AI vendor agreements — the precedent being set on procurement-as-policy will shape how funders, foundations, and implementing partners structure their own AI contracts.