Google DeepMind's AlphaProof Nexus solves nine open Erdős problems, a day after OpenAI's claim
Google DeepMind says its AlphaProof Nexus system generated machine-verified proofs for nine previously open Erdős problems, two of which had stood unsolved for 56 years, and also proved 44 open conjectures from the Online Encyclopedia of Integer Sequences. The system pairs a large language model with the Lean proof assistant, and each problem cost only a few hundred dollars to solve. The result landed a day after OpenAI claimed its own Erdős breakthrough. DeepMind was careful to say the system is still not AGI.
Why it matters: Verifiable, cheap machine-generated mathematics is moving from stunt to tool, and the head-to-head shows the frontier labs now competing directly on hard reasoning.
Anthropic is closing a $30 billion-plus round at a $900 billion valuation, set to pass OpenAI
Anthropic is finalising a funding round of more than $30 billion at a valuation above $900 billion, which would make it the world's most valuable private AI company, ahead of OpenAI. Sequoia, Dragoneer, Altimeter and Greenoaks are expected to co-lead, each investing roughly $2 billion, with the round set to close around the week of 26 May. Estimates put Anthropic's annualised revenue near $45 billion in May, up from about $9 billion at the end of 2025, with gross margin rising from 38% to over 70%.
Why it matters: The size of the raise and the margin shift signal that the leading model labs are now valued as durable infrastructure businesses, not research bets.
OpenAI publishes its Frontier Governance Framework
OpenAI released its Frontier Governance Framework on 28 May, a public document setting out how its safety practices map to new rules including California's Transparency in Frontier AI Act and the EU AI Act's code of practice. It covers risk assessment across cyber offence, CBRN, harmful manipulation and loss of control, and commits to external expert and third-party evaluation before a model crosses a new risk tier. It builds on the company's existing Preparedness Framework.
Why it matters: As frontier-AI laws come into force, labs are turning internal safety processes into public governance documents, setting the template regulators and enterprise buyers will expect.
OpenAI files a confidential S-1, targeting a Q4 listing near a $1 trillion valuation
OpenAI filed a confidential S-1 with the SEC on 22 May, starting the formal process to go public. It is targeting a Q4 2026 listing at a valuation between $852 billion and $1 trillion, with Goldman Sachs and Morgan Stanley leading the deal and a debut possible as early as September. The filing comes despite OpenAI losing about $1.22 for every $1 of revenue in Q1 2026.
Why it matters: An OpenAI listing at this scale would be the largest AI IPO yet and force unprecedented disclosure of frontier-lab economics.
HHS deploys ChatGPT to scrutinise state and grantee audits for fraud and waste
The US Department of Health and Human Services launched AERO, the Audit Enforcement and Risk Oversight initiative, on 21 May, using ChatGPT and other AI to ingest the annual single audits of any state or grantee spending more than $1 million in federal funds. Officials estimate $100 billion to $200 billion in wasteful or fraudulent spending a year and say the AI will flag chronic noncompliance, repeat deficiencies and delinquent audits. The penalty for noncompliance is loss of funding.
Why it matters: Many of those grantees are nonprofits, so AI-driven audit scrutiny raises the compliance bar for any organisation receiving significant federal money.
The National Lottery Community Fund commits £3m to charity-led AI
The National Lottery Community Fund announced a £3 million programme on 15 May to fund civil-society organisations to shape how AI is used, working with UK Community Foundations and the Centre for the Acceleration of Social Technology. It will run an "AI Pulse Network" pilot of 50 community organisations and back community-led AI tools rooted in local need, with one example being a benefits-advice charity funded to flag when an algorithm's decisions go wrong. First grants are expected in autumn 2026.
Why it matters: This funds charities to build and govern their own AI rather than only consume vendor tools, a model other funders may copy.
A new UK report finds charities using AI far faster than their boards realise
Charity Excellence's Charity AI 2026 report finds 76% of UK charities now use AI in some form, but only 3% of trustees believe their charity uses it at all, rising to 8% at larger charities. The report calls strategic awareness and governance "lagging" well behind frontline use. Separately, only 36% of UK adults said they feel positive about charities using AI, with 27% negative.
Why it matters: The gap between operational use and board oversight is a live governance and trust risk for nonprofits adopting AI without policy or disclosure.
AI Impact Hub