(Senior) Researcher - Misuse Risks

Key information 

Apply here until Sunday, 18 October 2026.

Role: Full time.

Location: In-person in the Bay Area or New York preferred. Remote considered for candidates in the US or UK.

Salary: 

• US: $122,052 - $172,056 + up to 3.5% 401(k) match
• UK: £81,462 - £120,579 + 10% unconditional pension contribution

If our posted range would meaningfully affect your decision to apply, we'd encourage you to apply anyway and flag it in your application. We’re happy to discuss compensation in more detail before an offer if it would affect your decision.

Ideal start date: As soon as possible.

About SAIF 

The Safe AI Forum (SAIF) is a US 501(c)(3) dedicated to advancing international cooperation to reduce extreme AI risks. We focus on fostering coordination, cooperation, and communication across the most important AI jurisdictions. Our flagship program, the International Dialogue on AI Safety, brings together senior computer scientists and AI governance experts from multiple jurisdictions, and we also run workshops, working groups, and a variety of other programs. Participants included Geoffrey Hinton, Yoshua Bengio, Stuart Russell, Andrew Yao, Zhang Ya-Qin, Craig Mundie, Fu Ying, and Xue Lan, alongside representatives from frontier developers globally.

We're a team of 16 based across the UK and North America. Over the next 12-18 months, the SAIF team will grow a lot, expand and launch new programs, and move into a new phase of our organisational history. We're impact-focused and move quickly, balancing rigour with speed and taking genuine ownership of outcomes. Our culture is warm and collaborative. People make time for each other and assume good intentions, even while working at pace on high-stakes problems.

About the role

Continued advancement of frontier AI systems will lower the barriers of malicious actors to launch societal-scale biological or cyber attacks. This is a sector where the leading AI jurisdictions, including the US and China, have overlapping interests. Preventing non-state actors from using AI to build weapons capable of mass destruction is in nation-states’ shared interests, and one of the few near-term possibilities for inter-state coordination. There is tractable work here: consensus on what a minimum safeguard standard looks like, shared understanding of how capable these systems actually are and how easy and effective terrorists may actually access dangerous capabilities, and practical mechanisms for developers on both sides to adopt and evaluate safeguards.

We are hiring one researcher to own and grow SAIF's portfolio in the intersection of AI safety and international diplomacy. You will run our programs on extreme misuse risks from frontier AI, such as hosting a dialogue that brings together Chinese and Western researchers and practitioners. This requires convening people who don't normally talk to each other, turning those conversations into outcomes, changing what developers and regulators do, scoping out and driving research / special projects to advance coordination and consensus-building, and building the internal picture that tells us where to push next. 

We are looking for someone who cares deeply about this issue and is ambitious about how far it can go. Within a few years, you would be one of the few researchers in the world who understand where extreme misuse risk and US-China coordination actually meet — and, as the work proves out, you will have the chance to build and lead a team around it.

What you'll do

Your main responsibilities: 

• Run the working-level dialogue end-to-end. You own the program: set the agenda and goals for each session, recruit participants and hold relationships across Western and Chinese institutions, facilitate the discussions themselves, and drive follow-through so momentum doesn't dissipate between convenings. 
• Convene people who don't usually talk to each other. Some of the highest-value discussions in this area put AI researchers in a room with counterterrorism, national-security, or biosecurity experts. You should be able to host those interdisciplinary conversations credibly — framing the problem so all participants can engage productively.
• Turn dialogue into consensus artifacts. Translate discussions into concrete outputs — shared frameworks, joint statements, best-practice documents — that give both ecosystems something to point to, adopt, and build on.
• Track the field, in both ecosystems. Maintain a best-in-class picture of the state of technical safeguards at Chinese and Western developers, how developers and policymakers on each side reason about open-weight model risks, and how the safety ecosystems are thinking about bio and cyber uplift. 
• Execute research projects. Scope and deliver high-quality technical governance work — for example, a practical safeguards playbook for developers on both sides — and commission or supervise external work where that gets to a good answer faster.
• Connect the work to policy. Coordinate with other SAIF researchers to translate consensus outputs into policy-relevant form, and test which deliverables actually land with policy stakeholders.
• Shape and drive SAIF's strategy in this space. You will be responsible for developing SAIF's strategy for facilitating US-China coordination on extreme misuse risks: spotting and scoping opportunities, testing bets, and communicating learnings and uncertainties transparently. In practice, you will likely start by producing regular internal memos that shape SAIF's direction — we value writing that is concise, concrete, and shows its reasoning. Over time, you will own a strategy for this area and drive it autonomously.
• Learn to navigate sensitive territory. Bio and cyber content carries political and legal risk, so you will work with SAIF’s broader to develop and refine internal policy in research and convening. 

You will report to Isabella Duan, who is leading a team at SAIF that focuses on building working- and senior- level coordination between Chinese and Western ecosystems on loss of control risks and extreme misuse risks. 

We work with an entrepreneurial mindset and believe in pivoting for impact — which means the content of your work may change. This is a new area we are testing in a fast-moving field, and if it doesn't prove out, we will work with you to pivot towards other priorities. 

Why you might be excited about this role

This role builds a rare combination: technical command of misuse risks and safeguards, the facilitation skill to run a room of stakeholders who don’t share assumptions, and genuine familiarity with how AI safety work is done in both Chinese and Western ecosystems.

Within a year you would:

• Understand the state of extreme misuse risks and technical mitigations in depth, in both ecosystems.
• Have trusted working relationships with researchers and practitioners on both sides.
• Have co-authored published consensus artifacts with researchers on both sides.
• Have real autonomy — running the programme with light supervision within twelve months, and the chance to build a team after that.

About you

Our ideal candidate is a technical governance specialist: in this role, you will not be running ML experiments or pushing the technical frontier, but you should be able to follow ML research closely enough to know what a safeguard actually does, and to productively engage with and facilitate technical conversations.

We care strongly about judgment, reliability, discretion, and commitment to the mission. We expect strong candidates will likely have most of the following:

Essential:

• Substantive grounding in AI safety or security. You understand how frontier systems fail, how misuse risk actually arises, and what current safeguards do and don't cover. This could come from technical research, red-teaming, a safety or policy team at a developer, or work at a safety institute or research organisation.
• Comfortable in policy discussions, not just technical ones. You can put AI researchers in a room with biosecurity or security specialists and keep the conversation productive — navigating sensitivities, finding the common ground that actually exists, and pushing participants toward shared understanding.
• Writing that decision-makers can act on. Clear, fast, and epistemically rigorous: you can compress a messy field into something a busy reader can use, and you are explicit about what you know, what you are inferring, and how confident you are.
• Evidence of owning something end to end. You have owned something end to end and been accountable for how it turned out — a program, a research project, a report.
• Comfortable with ambiguity. You can operate in a small organisation where the scope is not fully defined and the first version of everything is yours to build.
• An ownership mindset. You are proactive, you want to grow and own a space rather than execute someone else's plan, and you are excited by the prospect of leading a team as it grows.

Strongly preferred:

• Chinese-language capability. Reading technical and policy material without translation, and holding professional conversations in Chinese. 

Desirable:

• Existing familiarity with bio or cyber risk. Background in biosecurity, counterterrorism, cybersecurity, or nonproliferation. 
• Existing relationships in the Western or Chinese AI safety ecosystems, plus a working sense of how their institutions and incentives differ.

If you're excited about this work but unsure whether you're fully qualified, please err on the side of applying.

Other details 

• Start date: We expect to make an offer as soon as we find the right match, with a preferred start date as soon as possible. We understand that some candidates may need longer to exit existing commitments. If this timeline doesn’t work for you, please err on the side of applying anyway and let us know in your application in case we can make alternate arrangements.
• Benefits:
  • Unlimited PTO (no less than statutory entitlement in your jurisdiction, and minimum 20 days annually encouraged)
  • Comprehensive health insurance
  • Annual wellness budget of $6500
  • Annual childcare budget of $6500
  • Annual professional development support of $6500
  • Home-office setup and equipment budget of $1200
  • In-person team building events
• Working arrangements:
  • In-person in the Bay Area or New York preferred. Remote considered for candidates in the US or UK.
  • International travel required for team events and convenings (~5-8 times per year)
• Diversity:
  • We aim to employ people with many different experiences, perspectives, and backgrounds who share our mission. We are committed to creating an environment where all employees have the opportunity to succeed, regardless of race, religion, national origin, gender, sexual orientation, age, disability, pregnancy or maternity, marital or civil partnership status, or any other protected characteristic. 
• Requests for accommodation:
  • If you need assistance or an accommodation due to a disability, or have any other questions about applying, please contact careers@saif.org.

Application process

To apply, please fill in this application form by 11:59 PM local time on Sunday, Oct 18, 2026.

The application process will vary a bit depending on the candidate, but is likely to include written work and interview components, and a work trial. We offer payment for work samples and trials, conditional on your location.

If you think you might be a good fit for the role but are unsure whether you should apply, we strongly encourage you to do so.

If you have questions about this role, please contact careers@saif.org

Apply Here