Policy & Escalations Specialist

Substack•New York, NY

4d•$75,000 - $85,000

About The Position

Substack's mission is to build a new economic engine for culture. From the beginning, we have been clear in our belief in the free press and free speech—values we consider inseparable and fundamental to a trustworthy media system. To achieve our mission, Substack must uphold free expression while maintaining a vibrant platform with clear rules that allow for the open exchange of ideas. Substack's Standards & Enforcement team writes and enforces the rules that shape what happens on the platform. We're looking for a Policy Specialist who can turn messy, real-world abuse patterns into clear, defensible, and enforceable policy, and who reaches for data and AI tooling to do it faster and better than a purely qualitative approach would allow. This is a mix of policy, escalations and content review roles. You'll spend half of your time researching, drafting, and stress-testing policy. But you'll be expected to ground that work in evidence: how often does this behavior actually occur, who does it affect, what happens when we enforce, and what breaks when we don't. The content review part of this role ensures that you are always in tune with the evolving harm patterns in the platform and the escalations management part of the role ensures that the policy lines are consistently applied in high profile escalations.

Requirements

2–4 years working on content rules, Trust & Safety, or something close to it (law, regulatory work, investigative research). Experience actually writing policy is a plus.
Comfort with data. You can write SQL, work in a spreadsheet, and tell whether a number really supports the point someone is making. You don't need to be a data scientist. You do need to refuse to make a claim you can't back up.
Real experience with AI tools and curiosity to explore new solutions. You've used them for more than drafting text, you know where they tend to fail, and correct them when the output is wrong.
Strong, clear writing. The test is whether a reviewer can use what you wrote without asking you what you meant.
Good judgment when the answer isn't obvious, and the steadiness to make a call you can defend.
Willingness to look at upsetting content as part of core work. We provide support for this.

Nice To Haves

Familiarity with online safety regulation in the EU and UK. Experience turning regulatory requirements into operational guidelines.
Project management experience is a plus.
Experience with illegal content escalation workflows (NCMEC reporting, geoblocking decisions, LE request handling).

Responsibilities

Write and update our platform rules, plus the internal guidance our tools and reviewers use day to day; what counts, where the line is, what to do in the tricky cases.
Investigate new kinds of abuse as they show up (spam, scams, groups working together to game the platform, harassment, sensitive content) and recommend how our rules should respond.
Pull your own numbers to size up a problem and check whether our response is working, how common it is, how much we're acting on, how often people appeal and win.
Use AI tools to speed up policy and enforcement work; sorting review queues, summarizing long cases, spotting patterns across reports, and be willing to challenge those tools when they get it wrong.
Spend at least 50% of the time in queue review to get a hands-on understanding of current policies and the opportunities to improve them.
Develop the escalations management process and work with Legal, Product, Engineering, Support, and Comms on escalations that cross team lines.
Manage escalations involving illegal content categories with appropriate urgency, discretion, and reporting discipline.
Project manage the action items from different escalations and develop collateral for effective escalation management (ex. escalation playbook: intake criteria, SLAs, severity tiers, and handoff paths).
Handle the hardest cases, the ones that set a precedent, and write down your reasoning clearly enough that it holds up months later.