AI Safety Manager at Slingshot AI · LumiJobs

Responsibilities

Ash is a wellness app that gives people a place to talk through stress, anxiety, relationships or a bad day. It isn’t a replacement for therapy. People find it useful for everyday issues.
People bring deeply personal things to Ash, so new safety questions come up all the time, from user conversations, customer support, researchers and partners. We build our safety policy from what’s actually happening with our users and keep adding to it. We’re hiring someone to own this work full time: researching each issue, getting input from the right experts and deciding what Ash should do.
Safety work at Slingshot runs across teams. You’ll report to our General Counsel, who leads our regulatory work, and work closely with our machine learning, engineering, clinical and research teams to put our policies into practice.
Build and grow our policy book. Take each new issue and turn it into a clear working policy. Pull together the relevant regulation, professional guidance, clinical practice, ethics and research, then make a reasoned judgment call.
Bring in the right experts. Work with lawyers, psychiatrists, psychologists, therapists, philosophers and researchers, including our own clinical and research team and our advisors, and use their input to make practical decisions.
Work through hard trade-offs. Many safety questions pose conflicting goals, such as protecting other people against a user’s privacy, or steering someone toward a safer choice against respecting their autonomy. Help decide where we come down and set the process for making those calls.
Turn ambiguous issues into clear decisions. Investigate user reports, emerging risks and new requirements. Establish what we know, what remains uncertain and what we should do next.
Set priorities. Help the team decide which safety issues to tackle first, and give the people building safety features clear answers.
Turn policy into how Ash behaves. Each policy defines when it applies and what Ash should and shouldn’t do. Policies will feed directly into model training and into the guardrails, safe modes and banners that shape Ash’s responses, so they need to be clear and concrete, with examples of good and bad responses.
Make the response work in practice. Build clear processes for triage, investigation, escalation and follow-through. Define who does what, create the tools and documentation they need, and handle the operational details yourself.
Turn learning into product improvements. Work with engineering, machine learning, clinical and legal colleagues to translate findings into changes to Ash and the systems around it.
Build, test and improve quickly. For new systems, put a useful first version in place, learn from real cases and refine it. Match the pace and level of review to the consequences of the decision.

Qualification

You understand AIYou enjoy writing policyYou’ve worked professionallyYou’re open-mindedYou can handle a lot at onceYou learn fast and seek inputYou’re an excellent researcherYou’re resourceful and hands-on

Required

You understand AI. You know how people talk to AI, what might go wrong, and what controls a company does and doesn’t have (and should build). You don’t need to be a programmer. You might come from law, bioethics, philosophy, medicine, psychology, product management, engineering or another field.
You enjoy writing policy. You like pulling together many sources, from research and regulation to expert opinion, and deciding what our position should be.
You’ve worked professionally, ideally at a startup, where people build and do things, sometimes without a lot of guidance.
You’re open-minded. You don’t assume you already know the answers, and you can describe a time when evidence changed your approach.
You can handle a lot at once. You’ll work on many problems with many inputs, often at the same time.
You learn fast and seek input. You can get up to speed on an unfamiliar subject quickly. Your instinct is to get input from others before deciding, and then you make the call.
You’re an excellent researcher, writer and communicator. You use AI research tools well, and you can write clearly about complex questions, whether for a policy document or a public audience. You present well.
You’ve built something and made it better. It could be a product, an operational process, a program or an independent project. You can show how you moved from an unclear problem to something people used, then improved it through feedback.
You’re resourceful and hands-on. You’re willing to investigate individual cases, coordinate people, write the checklist and follow up until the work is done.
You handle sensitive information with care and stay thoughtful under pressure.
You care about people’s mental health and want more people to have access to good support.

AI Safety Manager

Responsibilities

Qualification

Required

Benefits

Company

Slingshot AI