AI news5
Stanford merges HAI and Stanford Data Science under James Landay
Stanford announced on 4 May 2026 that it is folding its Stanford Data Science initiative into Stanford HAI, with computer scientist James Landay as Denning Director of the merged institute. The combined body brings together more than 400 scholars, $60M in cumulative grant funding, the high-performance Marlowe GPU cluster, and HAI's industry affiliates programme. Founding director Fei-Fei Li takes a new role as Special Advisor on AI to Stanford President Jonathan Levin and co-chairs the advisory council with John Hennessy.
news.stanford.edu ↗Why it matters
HAI's AI Index is the most-cited annual cross-cut of AI adoption and policy data, and consolidation under a unified institute means the next Index will speak for the merged compute, scholarship, and societal-impact agenda in one voice.
OpenAI ships GPT-Realtime-2, Realtime-Translate, and Realtime-Whisper for voice apps
OpenAI released a refreshed family of realtime voice models in the API on 7 May 2026 covering smarter live voice reasoning (GPT-Realtime-2), real-time multilingual speech (Realtime-Translate), and streaming transcription (Realtime-Whisper). The models target voice-first product builders and replace the prior generation. Pricing and rate-limit details landed alongside the launch.
help.openai.com ↗Why it matters
Voice agents have been the slowest frontier for builders to ship reliably, and an integrated reasoning, translate, and transcribe stack from the default API drops the integration burden for everyone running a phone or kiosk product.
Salesforce overhauls Slackbot with 30+ AI features and deepens Anthropic Claude integration
Salesforce CEO Marc Benioff introduced what the company is calling its biggest Slack update since the acquisition, embedding more than 30 AI features into Slackbot. The release ships Claude Code in Slack via Anthropic, alongside agent integrations from OpenAI, Google, Perplexity, Writer, Dropbox, Notion, Cognition, Vercel, and Cursor. Slack is positioning the Slackbot AI surface as the new default place workers meet AI agents.
salesforce.com ↗Why it matters
The work-chat interface is becoming the default frontend for AI agents, which means platform choice (Slack vs Teams vs custom) is now an AI-procurement decision more than a comms one.
Anthropic ships Claude Code Auto Mode and ten Wall Street finance-agent templates
Anthropic released Claude Code Auto Mode and a new collection of ten ready-to-run finance-services agent templates the same week it doubled Pro, Max, Team, and seat-based Enterprise rate limits via the SpaceX Colossus compute deal. The finance templates cover pitchbook generation, KYC screening, and other workflows, built in collaboration with JPMorganChase and others. Claude Code added plugin URL loading, smarter package-manager auto updates, and gateway model discovery controls.
anthropic.com ↗Why it matters
Anthropic is moving from frontier-model vendor toward vertical-template provider, which compresses the build path for organisations that want bank-grade compliance workflow scaffolding without building it custom.
Snyk ships native Claude integration for AI-driven vulnerability discovery, prioritisation, and fixes
Snyk announced the week of 5 May 2026 a Claude-powered upgrade to its AI Security Platform that uses Anthropic's models to discover, prioritise, and remediate vulnerabilities across code, dependencies, containers, and AI-generated artefacts. The integration positions Claude as the reasoning layer behind Snyk's developer-ready fix recommendations. Coder, Opsera (Cursor plug-in), Prismatic, and Precisely also shipped agentic-security or DevSecOps tooling in the same week.
sdtimes.com ↗Why it matters
AI-accelerated coding has outrun traditional security review and the new wave of Claude-and-Cursor security tooling is the first credible attempt to put guardrails inside the IDE rather than after deployment.
AI in the nonprofit sector4
Microsoft Elevate + EY + Caribou Changemaker Fellowship for nonprofit AI adoption, applications close 22 May
Microsoft Elevate, EY, and Caribou opened applications on 22 April 2026 for the Changemaker Fellowship, a 10-week global programme moving nonprofit, UN, and intergovernmental staff from AI experimentation to responsible adoption. The first cohort onboards 24 June with an official kick-off on 29 June and takes about 25 fellows. The programme runs in 8 cohorts over 2 years and asks for 3-5 hours per week. Applications close 22 May 2026 at 23:59 PT.
ms-elevate.caribou.global ↗Why it matters
Free, named, and time-boxed AI-adoption fellowships from Microsoft and EY at this scale are becoming the standard credentialling pathway nonprofits will be expected to have on file when funders ask whether teams are AI-ready.
Claude for Nonprofits five-month metrics: MyFriendBen $1.2B routed, Epilepsy Foundation 24/7 to 3.4M Americans
Anthropic published updated impact metrics from Claude for Nonprofits, the GivingTuesday-partnered programme launched 2 December 2025. MyFriendBen achieved 4x faster implementation and connected families to over $1.2 billion in value across 70,000+ households. The Epilepsy Foundation runs 24/7 support for 3.4 million Americans through Claude. IDinsight reports 16x faster work on field research, and Global Health Partners built a geospatial dengue mapping tool in three days versus weeks.
anthropic.com ↗Why it matters
Five months in, the named-org dollar-and-population numbers are the clearest counter to the 2026 Nonprofit AI Adoption Report finding that 92% of nonprofits use AI but only 7% see major impact.
CareNX Fetosense and Spring ACT Sophia win AI for Good Summit awards for frontline impact
CareNX Innovations' Fetosense, a portable AI-powered fetal monitoring system, has been deployed in over 2,500 clinics across six countries and is credited with a 30% drop in NICU admissions, winning the AI for People Award at the AI for Good Global Summit in Geneva. Spring ACT's Sophia, a chatbot supporting domestic violence survivors in 172 countries with 24/7 anonymous guidance in 20+ languages, won the Pro Bono Collaboration Award. Both awards highlight production-grade nonprofit AI rather than pilots.
aiforgood.itu.int ↗Why it matters
Frontline-deployed nonprofit AI with hard outcome data (NICU admission reduction, 172-country reach) is the use case shape that wins funder attention in 2026, not chatbot demos.
Virtuous + Fundraising.AI 2026 Adoption Report: 92% of nonprofits use AI, 7% see major impact
The 2026 Nonprofit AI Adoption Report from Virtuous and Fundraising.AI surveyed the sector and found 92% of nonprofits use AI in some capacity, yet only 7% report major improvements in organisational capability. The remaining 85% sit in an efficiency plateau using AI for tactical wins without unlocking org-level transformation. Fundraising-specific impact, where AI was expected to outperform, lags relative to the cross-functional baseline.
prnewswire.com ↗Why it matters
A 92/7 split is the clearest data point yet for funders evaluating where to back AI capacity-building, and the report's framing of an efficiency plateau gives boards a vocabulary to push past surface-level adoption.