ERA:AI Fellowship Winter 2027: £10,000 Fully Funded Cambridge Research Fellowship in AI Safety and Governance

ERA:AI Fellowship Winter 2027: £10,000 Fully Funded Cambridge Research Fellowship in AI Safety and Governance

The Cambridge ERA:AI Fellowship is a 10-week, full-time research programme for people who want to work on technical AI safety, AI governance, or the technical side of AI governance. The Winter 2027 cohort is scheduled to run in Cambridge, United Kingdom, from 18 January through 26 March 2027. Applications close on 13 September 2026 at 11:59 p.m. Anywhere on Earth.

ERA describes the fellowship as talent-first. There is no formal academic or professional eligibility list beyond being at least 18 years old by the start of the programme. The programme is intended for a broad group of researchers and practitioners: technical specialists, policy and security professionals, lawyers, economists, students with strong project work, career changers, and people with distinctive expertise that could improve work on advanced AI risks. Formal research experience is helpful but is not required.

The financial package is unusually clear for a short research fellowship. Selected fellows receive a £10,000 stipend for the 10 weeks. ERA also says it covers travel to and from Cambridge, visa support, meals during working hours, and various social and research excursions. Its fellowship page also lists some research-related expense coverage, dedicated research management, a desk in the Cambridge office, and the possibility of additional support in special cases. ERA does not provide or guarantee accommodation, although the stipend includes a portion intended for housing and the team offers help with finding accommodation.

Key details

What the fellowship provides

The central benefit is protected time to complete a substantive research project. Fellows work full-time on a question related to reducing risks from advanced AI, with support from a mentor, an ERA research manager, and the wider fellowship community. The programme is designed for work that can be made concrete within 10 weeks. A successful project might produce a policy brief, working paper, technical report, replication or audit report, code with documentation and evaluations, or another output suited to the question.

The stipend is £10,000 for the 10-week programme. ERA’s FAQ states that the overall stipend includes a portion for housing, but the organisation does not provide or guarantee accommodation. That distinction matters when preparing a budget. Applicants should estimate Cambridge housing and living costs rather than assuming that a room is arranged by the programme. ERA says it can provide guidance and support to help fellows find accommodation together in Cambridge.

Travel to and from Cambridge is covered, and ERA supports visa applications for international fellows. The organisation also covers meals during working hours and various social and research excursions in and around Cambridge and London. The fellowship page mentions some research-related expense coverage, while the FAQ specifically describes the stipend, travel, visas, meals, and excursions. Applicants with unusual research costs should ask ERA what is covered before relying on an assumption.

The working environment is another important part of the package. Fellows use a co-working space in central Cambridge, have a dedicated desk, and work alongside other fellows and ERA staff. The programme typically includes speaker events, workshops, community activities, and opportunities to meet people working across research, policy, government, and industry. After the programme, fellows join an alumni network. ERA says it remains in contact with alumni and may provide career advice, opportunities, or recommendation letters.

ERA also says that top fellows may be considered for six-month-or-longer full-time funded positions through its Research Scholars programme. That is not an automatic continuation and should not be treated as part of the guaranteed award. It is best understood as a possible next step for particularly strong fellows whose work and interests fit a later ERA role.

Who is a good fit

The official eligibility rule is intentionally broad: anyone may apply if they will be at least 18 by the start of the fellowship. ERA explicitly welcomes people at any career stage and from many subject areas. It has accepted undergraduates with strong project work, graduate students, professionals moving from industry, and people with prior research careers. A conventional computer science or policy background is not required if an applicant can demonstrate the ability to think clearly about an important problem and contribute to the programme.

The strongest fit is likely to be someone who can connect a real area of knowledge to a tractable question about advanced AI safety or governance. For example, a technical researcher might bring expertise in evaluations, interpretability, robustness, alignment, control, or model monitoring. A policy applicant might focus on regulatory design, international coordination, corporate accountability, frontier safety frameworks, or governance of increasingly autonomous systems. A person in a cross-disciplinary area might propose work on compute governance, auditing, hardware-enabled verification, technical standards, or the connection between technical evidence and policy decisions.

ERA is also interested in distinctive perspectives. That can include experience in law, economics, security, public administration, social science, engineering, biology, or another field that changes how an AI safety or governance problem is understood. The key is not to force an existing background into fashionable terminology. Explain what you know, what question it lets you investigate, and why the question matters for the safe development or governance of advanced AI.

Applicants should also be realistic about the schedule. The programme runs in person in Cambridge for 10 weeks. ERA welcomes currently employed people, but accepted professionals need to arrange leave. Students need to arrange a break from their studies. Remote participation is not the default, although ERA says it may discuss exceptions with an accepted candidate in special circumstances. Treat attendance, housing, immigration, and leave as part of fit rather than as details to solve after applying.

Research streams and project scope

ERA lists three related streams. The Technical AI Safety stream includes work on evaluations, interpretability, robustness, alignment, control, dangerous capability measurement, model behaviour, safeguards, and human oversight. A good project in this stream should identify a specific technical uncertainty or failure mode and propose a method for producing useful evidence within the fellowship period.

The AI Governance stream covers questions about institutions and rules. Possible areas include regulatory design and implementation, frontier safety frameworks, corporate governance, international coordination, accountability, and policy options for autonomous AI agents. A governance proposal should identify the decision-maker or institution that could use the work, the evidence needed to support the recommendation, and the limits of the analysis.

The Technical AI Governance stream sits between the two. ERA describes interests including compute governance, auditing, monitoring, hardware-enabled verification, and technical standards. These projects ask how technical facts and infrastructure can make governance more precise and enforceable. A strong idea might examine how capabilities are measured, how a relevant claim could be independently verified, or how privacy and security constraints affect monitoring.

The three labels are not rigid boxes. ERA says it supports projects between established technical and policy areas. Applicants do not need to solve the entire problem in a short fellowship. A narrower project with a clear method, plausible evidence, and an achievable output is more useful than a grand proposal that cannot be completed in 10 weeks. Review the examples of previous research on the ERA site, then explain what your proposed work would add.

Application and selection process

The first stage is a written application with personal questions and short essay questions. ERA estimates that the written application should take about two hours. Applications are reviewed on a rolling basis until the deadline. The current deadline is 13 September 2026 at 11:59 p.m. Anywhere on Earth.

The next stage is a Round 1 interview. ERA says it expects to invite roughly the top 20 percent of candidates to interviews in late September. Applicants who advance will take part in a shorter Round 2 interview in early October. ERA will contact references for candidates in this later group. Final selection is holistic, and accepted fellows then work with ERA to develop project ideas and match with mentors.

ERA expects approximately 30 to 35 fellows for Winter 2027 across the three streams, with each stream typically containing 8 to 12 people. Those numbers are a planning estimate, not a guaranteed quota. The FAQ says the applicant pool and its quality can change the eventual distribution.

The organisation says it values reasoning transparency and epistemic humility. In practice, that means an application should show how the applicant thinks, not only present a confident conclusion. State what is known, identify important uncertainty, explain the assumptions behind the proposed approach, and say what evidence would change your mind. Showing the boundaries of an idea can make it more credible, especially in a field where both technical capabilities and policy conditions change quickly.

Materials and preparation strategy

Start by reading the official fellowship page and the FAQ, then use the Apply Now link from ERA’s site for the live form. The public pages identify a written application, but they do not publish a fixed checklist of attachments or a required proposal format. Do not invent a document requirement. Instead, prepare the information the written questions are likely to require: a concise account of your background, a clear research direction, evidence of serious work, and an explanation of why ERA is a useful setting for you.

For the project idea, prepare a one-page internal brief before opening the form. It should include the problem, why it matters, your proposed question, the method or evidence you would use, the likely output, and the main risks. This is not a claim that ERA requires a one-page brief. It is a practical way to make short essay answers specific and consistent.

Use evidence that a reviewer can inspect. Depending on your background, that may be a technical project, code, an evaluation, a policy memo, a published paper, a thesis chapter, a professional decision, or a carefully executed independent project. Explain your individual contribution and what you learned. ERA says formal research experience is not required, so an applicant without publications can still show research ability through a well-documented project and clear reasoning.

Make the timing credible. A 10-week plan should have a narrow first milestone, a method for checking progress, and an output that can be finished before the Research Symposium. ERA expects fellows to present their work and submit a written or otherwise suitable research output. You do not need to promise a journal publication. A rigorous policy brief, technical report, audit, or code-and-evaluations package may be a better fit if the question calls for it.

Finally, plan the practical side before submitting. Check passport validity, visa needs, leave or study arrangements, likely Cambridge housing costs, and your ability to work full-time in person. ERA says a passport must be valid until at least 26 September 2027 and have at least one blank visa page for the visa vignette. International applicants should allow time for any institution or employer letters they may need.

Common mistakes to avoid

Submitting a broad mission statement instead of a research question. Caring about AI safety is not the same as identifying a problem that can be investigated. Narrow the idea until the method and output are visible.

Overclaiming technical or policy certainty. ERA specifically values epistemic humility. Separate facts, assumptions, hypotheses, and recommendations. A thoughtful limitation is stronger than an unsupported prediction.

Treating every stream as interchangeable. Explain whether your project is primarily technical, governance-focused, or about technical infrastructure for governance. If it crosses streams, say what each side contributes.

Ignoring the in-person commitment. The stipend is not a substitute for an attendance plan. Budget for accommodation, arrange leave, and consider visa timing before the deadline.

Assuming accommodation or every expense is guaranteed. ERA says the stipend includes a housing portion, but it does not provide or guarantee accommodation. Ask the organisation about any unusual costs that your project may incur.

Writing for prestige instead of usefulness. The application should show what a decision-maker, researcher, evaluator, or institution could do with the result. Make the intended user of the work explicit.

Waiting until the final hours. The deadline is stated in Anywhere on Earth time, and applications are reviewed on a rolling basis. Submit before the final day if possible and keep a copy of the completed answers.

FAQ

Is the £10,000 a scholarship or a salary?

ERA describes it as a stipend for the 10-week fellowship. The public FAQ does not describe it as an employment salary. It also says that selected short-term extensions may include additional financial support, but extensions are not guaranteed.

Do I need a PhD or previous research job?

No formal research-experience requirement is published. ERA says prior research experience is helpful but not required and has accepted undergraduates, graduate students, industry professionals, and career changers. Your application still needs to show critical thinking, learning ability, and meaningful potential contribution.

Can I apply from outside the United Kingdom?

Yes. ERA welcomes talented people from around the world and says it sponsors visas and supports international participants with the visa process. You should still check your own immigration circumstances and passport validity.

Is housing included?

Accommodation is not provided or guaranteed. ERA says the overall stipend includes a portion for housing and that it can offer guidance and support with finding accommodation in Cambridge.

Can I participate remotely?

The default expectation is in-person participation in Cambridge. ERA says it may discuss remote involvement in some cases with an accepted candidate, but applicants should plan on attending in person.

What happens after the application deadline?

ERA says Round 1 interviews are planned for late September 2026 and Round 2 interviews for early October. References are contacted during the later stage, and final offers are followed by project development and mentor matching.

Official links and next steps

The official ERA:AI Fellowship page contains the Winter 2027 dates, funding, eligibility, research areas, and application process. The ERA Fellowship FAQ explains the stipend, accommodation, travel, visa support, attendance expectations, selection stages, and expected research outputs. Use the Apply Now link on either official page to open the current application form.

Before 13 September 2026, confirm that your proposed question fits one of ERA’s research streams, prepare concrete evidence of how you work, and decide whether you can commit to 10 weeks in Cambridge from 18 January through 26 March 2027. The programme is broad in who it can admit, but the application still needs a precise problem, a credible plan, and a clear reason that your perspective would improve work on AI safety or governance.