AI news5
OpenAI reasoning model disproves an 80-year-old Erdős geometry conjecture
OpenAI announced on 20 May that an internal general-purpose reasoning model independently disproved the planar unit distance conjecture, an open problem Paul Erdős posed in 1946. The model found a new infinite family of point arrangements producing at least n^(1+0.014) unit-distance pairs, beating the long-assumed grid bound, by routing through algebraic number theory rather than brute force. Princeton mathematician Will Sawin refined the result and Fields Medallist Tim Gowers called it a milestone for AI in mathematics.
openai.com ↗Why it matters
AI moving from solving known problems to producing genuinely new mathematics shifts what these systems can credibly be trusted to do in research-grade work.
Starbucks scraps its AI inventory-counting tool across North America after nine months
Starbucks retired its Automated Counting app this week, nine months after rolling it out across North American stores, according to an internal newsletter reported by Reuters. The tool was meant to flag shortages but frequently miscounted and mislabelled items, including confusing similar milk types, so staff return to manual counts. It was part of CEO Brian Niccol's push to fix the chain's persistent product shortages.
cnbc.com ↗Why it matters
A large, well-resourced AI deployment failing on a basic operational task is a useful reality check for any org tempted to automate before the data and workflow are solid.
Anthropic nears its first profitable quarter on the back of a SpaceX compute deal
Anthropic posted $4.8 billion in Q1 revenue and is on track for roughly $10.9 billion in Q2, which would deliver its first profitable quarter, per CNBC. A SpaceX compute contract worth $1.25 billion a month through 2029 includes reduced fees during a May to June ramp, cutting compute from 71 cents of every revenue dollar in Q1 to a projected 56 cents this quarter. For comparison, OpenAI reported about $5.7 billion in Q1 revenue, roughly $1 billion ahead of Anthropic.
cnbc.com ↗Why it matters
The economics of the lab most nonprofits build on are firming up, which matters for pricing stability and the odds it stays a long-term vendor.
Newsom signs first-of-its-kind California order to prepare workers for AI disruption
Governor Gavin Newsom signed an executive order on 21 May directing California agencies to develop policies for AI-driven workforce disruption. It tasks them with studying severance standards, transition support, worker-ownership models, universal basic capital, and even subsidies for companies that keep staff rather than replace them with AI. The order calls for recommendations within 180 days, including possible updates to the state's WARN Act on mass-layoff notice.
gov.ca.gov ↗Why it matters
State-level rules on how AI affects jobs will shape the policy environment that funders and social-sector employers operate in, well before any federal framework lands.
Spotify and Universal Music license fan-made AI covers and remixes
Spotify and Universal Music Group announced licensing agreements on 21 May to let fans create generative-AI covers and remixes of participating artists' songs, sold as a paid Premium add-on. Artists and songwriters share directly in the revenue, and the companies framed the model around consent, credit, and compensation. No launch date was given beyond the announcement.
newsroom.spotify.com ↗Why it matters
A major label and platform agreeing on paid, opt-in AI derivatives sets an early template for licensing creative work to AI rather than litigating it.
AI in the nonprofit sector4
Sentient Futures Summit opens in London today, focused on AI and animal welfare
The Sentient Futures Summit runs in London from today, 22 May, through 24 May, bringing together animal advocates, AI safety researchers, and consciousness researchers to discuss steering AI toward better outcomes for animals and other sentient beings. The Friday conference features expert-led sessions, with a weekend unconference for deeper collaboration and funder conversations. Organisers point to $2 to $4 million in funding being discussed across the AIxAnimals fellowship and project pipeline.
sentientfutures.ai ↗Why it matters
The intersection of AI and animal welfare is becoming an organised, funded field, directly relevant to AIH's animal-advocacy clients.
Microsoft Elevate Changemaker Fellowship closes applications today
The free 10-week Microsoft Elevate Changemaker Fellowship, run with EY and Caribou, closes applications today, Friday 22 May. It supports nonprofits on organisation-wide AI adoption through structured curriculum, expert coaching, and peer learning. This is the final-day reminder for any AIH-adjacent org still considering it.
microsoft.com ↗Why it matters
One of the few free, structured AI-adoption programmes for nonprofits this year, and the window shuts today.
IOM, UNHCR, and ICVA launch a dashboard mapping refugee protection gaps
The International Organization for Migration, UNHCR, and ICVA launched an online dashboard on 12 May mapping protection and assistance services for refugees and migrants, starting with the Western Africa Atlantic Route. It gives a cross-regional view of available services and urgent gaps to support referrals, programme design, and resource mobilisation.
iom.int ↗Why it matters
Shared data infrastructure across agencies is the unglamorous backbone that makes frontline AI and coordination work possible at scale.
Signpost deploys purpose-built AI agents for displaced people in Bangladesh and Nigeria
The IRC's Signpost programme, with aprendIA, is running purpose-built AI agents to help displaced people access food, education, and legal documentation, profiled this week by Rest of World. A teacher-support partnership in Nigeria's Borno, Adamawa, and Yobe states has grown from around 500 teachers to more than 4,700, and is expected to reach over 22,000 by the end of 2026.
restofworld.org ↗Why it matters
This is the concrete, measurable frontline AI deployment that funders ask for, with named tools, named regions, and real numbers.