AI news5
OpenAI launches Daybreak to counter Anthropic Mythos in the cyber defence race
OpenAI on Monday 11 May announced Daybreak, a cyber-defence platform that uses its frontier models to find, validate and remediate software vulnerabilities for enterprises and governments. The launch directly answers Anthropic's Mythos (Project Glasswing). Daybreak builds on OpenAI's Trusted Access for Cyber (TAC) program, which now spans hundreds of organisations and thousands of individual defenders.
openai.com ↗Why it matters
AI-native cyber defence is the first frontier-model application area where the two leading labs are publicly head-to-head with explicit product framings, meaning enterprise security spending will start flowing to whichever vendor wins the proof-of-deployment race.
OpenAI agrees to cap Microsoft revenue share at $38B
The Information reported on Monday 12 May that OpenAI has agreed to cap the total revenue it shares with Microsoft under their existing agreement at $38 billion. The cap is a hard ceiling rather than a percentage cut, materially reshaping the financial relationship as OpenAI pushes through its for-profit transition. The reporting comes a day after the OpenAI Deployment Company launch.
reuters.com ↗Why it matters
Microsoft has been OpenAI's distribution backbone for three years. A fixed cap means OpenAI's incremental enterprise revenue past that threshold flows straight to its own balance sheet, accelerating the strategic decoupling that the OpenAI Deployment Company already signalled.
OpenAI publishes Parameter Golf takeaways: coding agents reshape ML talent discovery
On 12 May OpenAI published learnings from Parameter Golf, an open ML challenge with a 16MB model artifact limit and 10-minute 8×H100 training budget. Over eight weeks the contest received 2,000+ submissions from 1,000+ participants. OpenAI explicitly frames the contest as a talent discovery surface and notes that AI coding agents materially lowered the cost of experimentation and broadened participation.
openai.com ↗Why it matters
For the first time, a frontier lab is publicly treating an open competition as both a research output and a recruiting funnel calibrated to AI-agent-assisted work. This is what the next layer of ML hiring looks like.
Google disrupts hackers using AI to weaponise zero-day vulnerabilities
Google announced on Tuesday 12 May that it had disrupted a criminal operation using AI to discover and exploit a previously unknown zero-day vulnerability in a corporate target's defences. The exploit would have bypassed signature-based controls. Google attributed the disruption to its own AI-powered defensive systems running in parallel with the attackers.
fortune.com ↗Why it matters
This is one of the first confirmed real-world cases of AI-driven offensive cyber actually being countered by AI-driven defensive cyber in production, validating the Daybreak and Mythos commercial pitches that landed this week.
Microsoft maps global AI adoption in new "State of Global AI Diffusion 2026" report
Microsoft On The Issues published "The State of Global AI Diffusion in 2026" on 7 May. The report visualises AI usage intensity by country and identifies a widening gap between high-uptake markets like the US, UK, India and Israel and slower-adopting economies. It positions Microsoft Elevate as the company's lever to close that gap. The report is now Microsoft's primary policy talking point in trade engagements.
blogs.microsoft.com ↗Why it matters
For Australia and ANZ specifically, this report shapes how Microsoft frames its $25B AU investment and Elevate-for-Changemakers rollout. Expect to see this data cited in every AU government and NFP AI conversation through 2026.
AI in the nonprofit sector5
NFP Digital Futures launches with $16M to lift Australia's not-for-profit sector
Infoxchange and PwC Australia's Impact Assembly launched NFP Digital Futures on Monday 11 May at the Technology for Social Justice Conference in Melbourne. The initiative has $16 million in early funding secured with an ambition to mobilise $50 million, backed by Greater Melbourne Foundation, Paul Ramsay Foundation, Ian Potter Foundation ($1.25M over five years), Minderoo Foundation, Gandel Foundation, the National AI Centre, Microsoft, PwC and Okta. It coordinates 35+ NFP, government and industry organisations including ACOSS, ACNC, SVA, Smith Family, Alannah and Madeline Foundation and Justice Connect. Focus areas are digital foundations, cyber security, data capability and responsible AI adoption.
infoxchange.org ↗Why it matters
This is the single largest cross-sector NFP digital capability play in Australian history, and responsible AI adoption sits inside the founding scope. For any AU AI-for-nonprofits operator, the next 18 months of sector funding flow through this coalition.
Code for America and Anthropic ship SNAP Policy Navigator for caseworkers
Code for America and Anthropic announced a partnership at the 2026 Code for America Summit in Chicago on 8 May to build Claude-powered AI tools for government caseworkers. The first product, SNAP Policy Navigator, gives caseworkers real-time access to verified federal, state and county SNAP policy guidance while processing individual cases. CEO Amanda Renteria framed the partnership as "responsible AI capable of easing caseworker burden, streamlining processing, and delivering benefits more quickly and accurately." The pair plans to scale reusable tools across states.
codeforamerica.org ↗Why it matters
This is the first production-grade Anthropic deployment inside US public benefits administration with named outcomes. It sets a procurement template that other state and federal agencies will copy.
Tipping Point Community commits $40.3M to AI workforce resilience in the Bay Area
At its annual gala on 8 May, San Francisco anti-poverty nonprofit Tipping Point Community announced a $40.3 million commitment to prepare low-income Bay Area workers for an AI-driven economy. The fund was anchored by a $25 million gift from the Sobrato family. CEO Sam Cobbs unveiled the strategic pivot from broad anti-poverty work toward AI reskilling. Tipping Point is the former pet project of San Francisco mayor Daniel Lurie.
sfstandard.com ↗Why it matters
A flagship Bay Area anti-poverty nonprofit reorienting its entire grantmaking thesis around AI workforce displacement is the clearest signal yet that AI economic disruption is being treated as the new poverty frontline by major US donors.
Partnership on AI receives $500K from Humanity AI to accelerate sector work
Partnership on AI (PAI) announced today, 12 May, that it has received $500,000 in funding from Humanity AI, the collaborative philanthropic initiative formed late last year to ensure AI benefits all of humanity. The grant supports PAI's work across AI safety, AI and labour, and AI and human connection programs. PAI is the longest-running multi-stakeholder AI body, with members across labs, civil society and academia.
partnershiponai.org ↗Why it matters
Humanity AI's first wave of grants is now visible. Watch where the next several land, because they will signal which AI civil-society institutions the collaborative considers worth scaling and which it bypasses.
Agog Foundation opens $6.5M immersive media for good fund with up to $1M grants
Agog announced on 12 May a new $6.5 million call for proposals supporting creators and nonprofits using immersive media (XR, spatial, AI-generated 3D) to drive climate action and civic engagement. Individual grants run up to $1 million. The fund explicitly bridges AI-generated content with nonprofit storytelling and civic media.
tmcnet.com ↗Why it matters
Immersive media + AI is the next frontier for nonprofit storytelling beyond the static-asset and short-video formats most NFPs default to, and $1M grants are large enough to underwrite serious production rather than pilots.