Cambridge ERA:AI fellowship 2027 pays £10,000 for AI safety research — apply by 13 September

Applications for the Cambridge ERA:AI Winter 2027 Fellowship are open on the official ERA:AI application page. According to the official announcement, the programme is hosted in Cambridge, UK, runs for 10 weeks from 18 January to 26 March 2027 and accepts a deliberately broad group: technical AI safety researchers, AI governance professionals and technical policy analysts — anyone aged 18 or older, from any academic or professional background, anywhere in the world.

What the fellowship actually is

ERA:AI is not a course in the traditional sense. Fellows receive a £10,000 stipend, separate from the programme’s in-kind benefits, and are expected to produce research, not just sit through lectures. The official announcement lists mentorship, office space, research support, networking, visa support where eligible, meals during working hours and eligible travel expenses as benefits alongside the stipend. The fellowship supports research across three tracks:

• Technical AI safety — the hard problem of making models behave reliably under deployment pressure.
• AI governance — how institutions oversee AI development, from lab practices to public policy.
• Technical AI policy — the intersection, where policy is written by people who actually understand the systems.

ERA:AI frames that technical policy profile as a genuine gap: the track exists for people who can move between policy language and model behaviour. Whether a particular compliance team has enough technical literacy varies by organisation, but the AI Act makes that question more visible for systems within Article 50’s scope.

The money: £10,000 in context

Let us do the arithmetic. The £10,000 stipend works out to about £1,000 a week for the ten-week programme.

Cambridge is not a cheap city. The stipend covers a modest student budget for the duration, but if you are coming from outside the UK, check accommodation costs before applying. The official announcement distinguishes the £10,000 stipend from the in-kind benefits — visa support where eligible, meals during working hours and eligible travel expenses — so budget for housing separately.

Europeans get an extra day — but only one

The application deadline is 13 September 2026 at 11:59 PM Anywhere on Earth (AoE). AoE runs 12 hours behind UTC, which means most of Europe gets what looks like a deadline extension.

In Central European Summer Time, the real cut-off is 14 September at 13:59. In the UK it is 12:59 BST on 14 September. Do not be the applicant who reads "23:59 AoE" and assumes it means midnight in their own time zone. For Prague, Berlin, Paris or Rome, the deadline is Monday afternoon, 14 September.

Why this round matters: the AI Act is in enforcement mode

This application cycle lands at a specific regulatory moment. Under Article 50 of the EU AI Act, the relevant transparency obligations — such as labelling certain synthetic content and disclosing certain chatbots as AI — apply from 2 August 2026 to covered systems, with enforcement by the European AI Office and national authorities, according to the European Commission’s AI Act implementation guidance.

The practical consequence for organisations whose systems fall within the scope of Article 50 is that they may need people who can do AI governance with technical credibility. Someone who can read a model card, assess whether safety claims hold up in practice, and determine which specific Article 50 transparency obligations actually apply to a given deployment — that profile is exactly what ERA:AI’s governance and technical policy tracks are designed to support.

Cambridge sits on the UK side of post-Brexit Europe, but the programme is globally open. International applicants may be eligible for visa support and should confirm UK entry requirements with the organisers. For European researchers who want to stay in the AI safety ecosystem without relocating to California, the fellowship is a funded route with a direct link to EU compliance work. Organisations should assess whether the Article 50 transparency obligations apply to their specific use case before treating them as mandatory.

After the 10 weeks

High-performing fellows may have a potential pathway to an additional six months of funded research through the Research Scholars programme. The organisers do not present this as a guaranteed extension; it is a separate opportunity for strong participants. If selected, that six-month research period can act as a potential career entry point for someone pivoting into AI safety or governance, rather than a gap-filler.

Should you apply?

If you are 18 or older and have a demonstrable interest in AI safety, governance, or technical policy — yes. You do not need a machine learning PhD; the programme explicitly welcomes applicants from any background. What you need is a question or project strong enough to spend 10 weeks on, plus a willingness to be in Cambridge in January.

The window is short: 13 September, 23:59 AoE, which is 14 September at 13:59 in Central Europe and 12:59 BST in the UK. For a funded research programme with no degree barrier, that is not much time. Set your calendar now.