Head of Guidance and Preparedness

Most high-stakes moments in AI over the coming years will involve frontier company insiders having and raising concerns: within companies, to regulators, or through other pathways.

Those insiders tell us they face enormous uncertainty along the “concern raising” path -- from detecting concerns to whether and how to voice concerns with maximum impact and minimal risk, which even legally protected whistleblowing can carry.

To date, no organization has dedicated themselves deeply to delivering the best strategic, frontier-AI-specific guidance to these individuals, their supporters, and their attorneys (our partners). We believe the ecosystem surrounding insider support is currently unprepared to handle disclosures optimally and at speed. This means critical disclosures may not occur at all or sub-optimally. As these moments may be critically high-leverage, making them “go well” has dramatically outsized impact.

Making “disclosures go well” requires (broadly*) two things: 1) Having the best guidance to insiders available and 2) ensuring that this guidance reaches those who need it.

This role is responsible for 1), and will provide input into 2), where execution will be owned by our Education Delivery Team and our partner network.

This is the most critical role we have hired for at AIWI to date as we scale by ~1 OOM over the coming year. It comes with a high 6 digit annual budget, headcount responsibility and significant latitude to achieve the goal outlined in 1) above. It unites strategic, long-term thinking with highly pragmatic and “rapid response” work.

*It also requires improved legal protections, disclosure channels, security -- which other AIWI teams handle and which are not core to this role.

About the role

AI company insiders experience dramatic uncertainty as missteps can be highly consequential. The “guidance gap” is most pronounced on strategic and AI-specific considerations and not, for instance, on legal matters or journalist-interactions, which our network can close.

Your job is to find answers to these questions:

1. Where should we focus? Which disclosure scenarios do we believe are most critical, unlikely to occur optimally (safe, net-positive impact), and tractable to improve?
2. What guidance do/ will insiders need in these scenarios? What should ‘trigger alarms’? How can one verify a concern?
3. What do we do when it happens? Who should be called, what should be asked, and what does a robust response look like for disclosures in our highest-priority risk categories? What pitfalls do we have to avoid, and how can we steer disclosures into robustly positive territory?
4. What should we build to move the needle? Based on our insights, which infrastructure should we/ our community build? For which scenario should we be prepared?

No one in the ecosystem has built the processes to answer these questions rigorously. This means we believe the world is currently insufficiently prepared for those critical moments in our future. You close this gap.

What you will do

To find answers to the questions above, we imagine you would want to conduct work such as the following:

Design and run research to uncover guidance gaps and close them. We expect you to largely do this through stakeholder consultations and scenario workshops across priority risk categories. Commission and manage fellows and contractors. You may also want to design (and publish) automations in problem discovery, research, and network maintenance, supported by our Tech team. Examples of research questions:

• Which risks relating to AI are only disclosable through insiders, and how do these relate to external verifiability?
• What do signals for these risks look like? How should disclosure thresholds differ across risk categories --- loss of control, misuse, power concentration --- or throughout functions --- model behaviour, governance failures --- differ? How does the calculus shift when evidence is ambiguous? When does accumulated cultural drift cross a threshold -- and can that threshold be made actionable for insiders experiencing it incrementally?
• Which teams — safety policy, internal deployment, post-deployment monitoring, internal red-teaming, global affairs — sit at the intersection of high-risk exposure and low external visibility? Which individuals must be educated?
• Through which mechanisms could disclosures be “net-negative”? How can we reduce “false positive” vs. “false negative” trade-offs?
• How can we maintain up to date models of company-internal power and decision making dynamics? What theory-of-change development support is valuable for us to do?
• How do shifting timelines impact considerations around where those thresholds are? Which timelines/ scenarios do we want to focus on (if decisions are required)?
• Are there evidence signatures that reliably accompany critical disclosures? Is the current ecosystem positioned to detect them, and what would need to be built if not?
• Practical: How can insiders safely build coalitions around a concern? What should insiders do who considering resigning over a concern? How can insiders detect motivated reasoning in themselves and others? What signals should insiders “look out for”?

Convert insights into actionable processes and guidance for insiders, partner attorneys, and support organizations. Together with our Education Delivery function, train insiders and partner organizations on our thinking. Guidance may include processes for partner attorneys to use, publications, PDF guides, interactive tools, presentations, etc.

Build expert network and intelligence. Maintain a trusted pool of experts you can call on rapidly and who will take your call. We should know who to call on what.

Build preparedness plans for when a disclosure occurs. Prepare optimal responses for high-sensitivity scenarios to respond optimally within hours. This may include having verification questions/ contacts set-up, journalists pre-briefed, open letters pre-populated, etc. This thinking directly feeds into the guidance/ strategy work.

Shape AIWI's strategy. Advise on which risk categories, company functions, and disclosure types AIWI should prioritize. Initiate new project to close major gaps in ‘whistleblowing chain’, e.g. allowing employees to coordinate safely for disclosures. Work closely with our policy team.

Your profile

A variety of profiles and backgrounds may fit this role, and the role will be shaped to and by the right person. We expect a successful individual to have multiple of the following traits:

• You understand the weight of this role and display unwavering integrity. Your thinking and actions will shape whether critical safety concerns, shaping our joint trajectory, are disclosed optimally and what it costs the person(s) who do so. This should weigh on and motivate you.
• You are relentlessly resourceful. You decide what needs to happen and make it happen. You get around obstacles instead of stopping at them. You like hard problems with no, initially, clear answer. You are not afraid of “getting your hands dirty”.
• People already approach you for advice today. In your professional and private life, people approach you for and often end up following your advice in difficult situations. Ideally, you have counseled individuals working at AI companies in some capacity before.
• Extremely strong strategic thinking and opinionated judgment under uncertainty. You are not overwhelmed by complexity. You prioritize consistently, think in hypotheses to be tested, can rapidly distil core considerations and synthesize input from specialists, rapidly forming well-calibrated, pragmatic and actionable opinions. Your thinking is non-dogmatic and integrates uncertainty (e.g. on timelines or risk). You take seriously a variety of risks.
• You have a track record of independently managing projects that successfully tackled large, complex challenges. Including recruiting and managing collaborators, supporters, and various stakeholders.
• Past employment at a frontier company. Understanding organizational practices in frontier AI development firsthand. Critical disclosures will largely be shaped by organizational behavior. You ideally have first hand experience working for at least one frontier company in a role that allowed you to witness high-context decision making. At a minimum, you should be highly interested in organizational dynamics.
• Technical understanding. You should be able to follow a description of a technical issue and be able to ask the right questions to assess the degree of concern.
• Epistemic humility: You know when to bring in more voices and are aggressive in seeking out feedback to ensure our thinking is appropriate to the challenges we face.
• Evidence of ability to communicate with non-technical audiences. You could brief a regulator on which scenarios are most concerning to us and draft a threshold decision a partner attorney can use with a client on the same day.
• You bring a relevant, trusted network.

You do not need a whistleblowing/legal/security background; this will be provided to you by our partners/team members/ experts you will bring in. You will however have to be deeply interested in rapidly gaining an understanding of those domains.

Backgrounds likely to be suitable may include:

• Safety or policy lead at a frontier company
• Senior researcher at an AI evaluation or governance organisation – likely with a focus on incident response, emergency preparedness, threat modelling, or comparable field
• Senior leader at an AI advocacy organization
• Senior grantmaker with deep AI portfolio knowledge

Why join?

• Impact: Influence critical moments in humanity's shared trajectory. Help individual humans in need of support.
• Intellectually stimulating: Think deeply about intellectually demanding questions with individuals who lead thinking in their fields. Unite strategic and pragmatic thinking.
• Independence & ownership: Build your team, manage your own budget in an environment that celebrates experimentation and execution.
• Growth: Shape AIWI as we scale dramatically as an organization.
• A supportive team: Who will stand by your side.
• Competitive compensation and runway secured until 2029

Who you’d be joining

AIWI is an independent non-profit launched in late 2024. We are ~4 FTE today and are scaling to ~15 FTE within the year, spread across the Bay Area, DC, and London. Joining now, you will significantly shape what the future of AI insider support looks like. We are funded until 2029.

Examples of our work to date:

• Policy: We led to the creation of the world's first AI whistleblowing channel. We pushed for and shaped the EU AI Office's channel. Ask us about our US work.
• Insider education & support: Our online campaigns have reached thousands of company employees. We produce guides (available upon request), online courses, and help establish connections to our great partner attorneys/ support (e.g. on legal advice/ defense funding).
• Advocacy: Anthropic published its whistleblowing policy and OpenAI strengthened theirs because of our Publish Your Policies Campaign.
• Network: We closely collaborate with the world’s leading whistleblowing and AI organizations.

Our ambition is to ensure each disclosure reaches the person best able to act on it and that they understand what’s at stake, all while maximising safety for the person making the disclosure. Joining us means helping us achieve that goal.