A note for people who run things

AI safety's bottleneck isn't researchers anymore. It's operators.

If you lead communications, people, recruiting, operations, finance, or strategy, the organizations working on how AI development goes need you more than they need another ML PhD — and most of them are bad at telling you so.

Here's what's actually true about the work, and where to look.

The short version

• These are real organizations with real budgets. Research labs, evaluation and standards bodies, national AI safety and security institutes, policy shops. Many have gone from a dozen people to 50–200+ in three years, funded by philanthropy, government, and revenue.
• They built research first and the organization second. Most were founded by researchers. A good number are now running a 100-person institution on the operating spine of a 15-person one — no head of people, no recruiting function, comms handled by whoever writes fastest. That gap is the bottleneck, and it's getting expensive for them.
• The open roles are your roles. Not adjacent to your job — your job, at an organization where it happens to matter unusually much.

• Head of Communications
• Head of People
• Recruiter / Talent lead
• Chief of Staff
• Head of Finance
• Operations lead
• Program & events management
• Government relations
• Development / grants
• Strategy
• Legal & compliance
• Product

• The compensation is real. Senior operator roles at the better-funded organizations commonly post in the $150,000–$260,000 range, and most listings show the band up front. It varies a lot — small nonprofits pay less, labs pay more — but this is not a vow of poverty, and you should stop assuming it is.
• You do not need to be technical. You need to be excellent at your function and willing to learn the subject matter. The subject matter is learnable in months. Twelve years of running a communications function is not.

The money is already committed. The people aren't.

If you want one fact to decide whether this is a real market or a nice idea, it's this one.

In July 2026, a new research organization called Resolution announced a $160 million grant from Coefficient Giving — the largest that funder has made in this area. The structure is the interesting part: $108 million base, plus $52 million conditional on a combination of hiring success and compute needs.

Read that again as an operator would. A funder wrote a nine-figure cheque and made a third of it contingent on whether the organization can actually staff itself. That is a hiring problem priced by someone with money at stake — which is a considerably stronger claim than anyone's opinion, including the rest of this page.

Resolution launched in June 2026 with seven founders and is "aiming at 40–80 FTE two years from now," hiring "across research, engineering, operations, and security," out of Berkeley plus remote in London and Melbourne. Seven people to eighty in two years is not a research problem. It's a recruiting, people-ops, and finance problem, and it is the same problem at a dozen organizations at once.

Nor is that one grant an outlier. Coefficient Giving — the funder behind it — says it expects to move "around $1 billion" in catastrophic-risk grants in 2026, has been staffing up its own grantmaking bench, and describes the push as an effort to "encourage our grantees to raise their ambitions." Grantees raising their ambitions is, in practice, grantees hiring.

Money moves faster than hiring does, and right now the gap between the two is the opportunity. If you have run a function before, you are on the short side of that trade.

Three assumptions that are doing most of the blocking

"They only want ML PhDs."

This was closer to true in 2019. It isn't now. 80,000 Hours — which researches careers for a living, and is not selling anything either — published "AI safety needs more than engineers" in April 2026 making exactly this point: these organizations "hire people, run events, raise money, and share their ideas with the outside world," and they are short of the people who do that well.

"I'd be starting over."

The opposite is closer to the truth: your function is scarcer inside these organizations than it is in your current industry. A head of comms who has handled a genuine crisis, a recruiter who has actually closed hard senior searches, a chief of staff who has scaled a company through 100 people — those are rarer here than in tech, and more consequential. "How to pivot without restarting your career" is the better version of this argument.

"It's not a serious career move."

People with serious careers keep making it. Three, profiled by 80,000 Hours in June 2026:

• 10+ years running humanitarian crisis communications at the UN and the Red Cross → communications strategy lead at SaferAI
• ~10 years as a product manager in finance, building analytics for investment banks → research manager at the UK AI Security Institute
• five years in quantitative portfolio management, ~$20B in funds and ETFs → head of finance at Goodfire

None of them retrained as engineers. All three took months, not weeks — a real search, with real rejections.

Where to look

There are already good boards for this. This page isn't trying to be one — it's a signpost to the people who do it properly.

80,000 Hours job board → The broadest curated catalog. Filter by role type — operations, communications, management, policy — rather than browsing everything. AISafety.com/jobs → Community-maintained aggregator, several hundred live roles, with skill filters for Operations, Outreach, and Management. Good complement — it catches postings the others miss. AI Safety Careers → A curated board covering safety, governance and policy roles, updated near-daily, with salary bands where employers publish them. Successif — talk to a person → Free one-to-one advising built specifically for mid-career and senior professionals: someone reads your background, tells you which roles you'd actually be competitive for, and makes introductions. If you only do one thing on this page, do this one — it's the fastest way to find out whether this is real for you.

The honest caveats

• Many of these are small organizations. You may be the first person to hold your function, with no team under you for a year. If your excellence depends on having a bench, read the org chart before you fall in love.
• Location varies more than you'd like. A lot of roles are remote-friendly; a meaningful number want you in San Francisco, London, or Washington. The listing will say. Believe it.
• The field argues with itself in public. There are real disagreements about what matters and how fast. You'd be joining a contested space, not a settled one — which is part of why judgment from outside is worth so much here.
• Not every organization is well run. Some are excellent; some are three researchers and a mission statement. Diligence them the way you'd diligence a Series A you were about to join: who's on the leadership team, what's the funding runway, who held this role before and why did they leave.

If you're in the small group this is for

This isn't a pitch for everyone, and it isn't a charity ask. It's a market observation: a set of well-funded, high-stakes organizations are short of exactly the skills you spent your career building, most of the people who have those skills have never heard the opportunity described in their own professional language, and so the roles sit open longer than they should.

For a small number of unusually capable operators, that's the highest-leverage move available right now. If that might be you, the links above are the whole ask.