Opportunities 2026-08-09

Eight weeks, $12,000 and a GPU budget to work on AI safety — from anywhere, if you apply by 17 August

The AI Alignment Foundation, a US non-profit, is taking applications for its 2026 fellowship, run with AE Studio, whose alignment team has published research with Anthropic. The programme runs eight weeks from 8 September to the end of October, full-time at forty hours a week, entirely remote and open to applicants working from anywhere. It pays a $12,000 stipend in two instalments, and each team gets cloud GPU access, API credits and a research manager who works with them daily on what to prioritise and how to track experiments. It closes with a Demo Day in Los Angeles; travel is covered for those who attend in person, but the foundation does not sponsor visas, so anyone needing travel documents has to arrange them alone. Applications ask for a CV, a questionnaire and a video of no more than five minutes answering two prompts — why alignment matters, and where you think the foundation's own research directions are wrong. The bar is technical: organisers want a strong machine-learning background, prior independent research and the ability to work without close supervision. They say plainly that this is not the right programme for people who need substantial guidance, or whose interest is in AI policy rather than technical work. The deadline is 17 August, anywhere on earth.

Why it mattersPaid, remote and open worldwide is a rare combination in this field. Most routes into safety research still run through a handful of labs in San Francisco and London and assume you can move there, which quietly excludes people who cannot — and $12,000 for eight weeks plus compute is enough to make the work possible rather than merely permitted. The compute matters more than the money for anyone who has tried to run real experiments on a personal budget. Two cautions before you spend a weekend on the application. The eligibility bar is genuine, not decorative: this is aimed at people who have already done research on their own, not at newcomers hoping to be taught. And the no-visa-sponsorship line is the one that catches people out — the research is remote and that part is genuinely open to anyone, but the closing event in Los Angeles is not, unless you can already get yourself into the United States.

✓ Verified · 2 sources

Read in the app — free, in 9 languages

A free year of Google's paid AI plan is open to college students in 140-odd countries — and it has to be claimed by 31 December
2026-08-20 A 90-day agent-building contest opened on Friday: open to anyone anywhere, free to enter, ₹18 lakh in prizes, closes 15 November
2026-08-17 A paid eight-week alignment fellowship closes on Monday — and it asks for no institution behind you
2026-08-15 Two weeks left: Google's Accra lab offers African founders DeepMind models before anyone else sees them
2026-08-14 OpenAI offered 100,000 researchers free frontier access. The first round has become a lottery.
2026-08-12

Eight weeks, $12,000 and a GPU budget to work on AI safety — from anywhere, if you apply by 17 August

Related stories