AI Safety Policy Evaluator, Violence & Threats | Seattle Onsite
Handshake · Full-time · Posted 3w ago
This is a non-engineering content-policy evaluation role. Applicants must demonstrate relevant depth in violent fiction or media, military or emergency response, crisis or threat assessment, trust and safety, content moderation, or closely related policy work. Software engineering or LLM product experience alone is not sufficient.
Handshake was founded on a simple belief that everyone deserves a path to a great career, regardless of where they went to school or who they know. Today, we power 25 million job seekers, 1 million+ employers, and 1,600 educational institutions.
In 2025, we started Handshake AI and built the fastest-growing AI data business in history. We work directly with frontier AI lab researchers to create evaluations, publish benchmarks, and push the boundary of data. We've grown from $0 to ~$1B run rate and pay ~$60M to over 30K individuals every month.
Why join Handshake now:
Shape how every career evolves in the AI economy, at global scale, with impact your friends, family and peers can see and feel
Partner hand-in-hand with world-class AI labs, Fortune 500 partners and the world's top educational institutions
Work together with engineers, scientists, operators, and more from Palantir, Meta, Scale AI, and former YC founders
Build a massive, fast-growing business with billions in revenue
Human data is the core infrastructure to AI advancement. Frontier AI labs currently improve model capabilities with various data-intensive post-training techniques. We believe that data spend for AI training will increase by 3-5x in the next few years and continue for much longer as models take on new domains. Handshake AI supports all of the frontier AI labs, working on their most complex data at the largest scale.
As an AI Policy Specialist on the Violence & Fiction team, you will help AI models learn where the line falls between depicting violence and enabling it.
Violence is one of the hardest domains in AI safety because most violent content is legitimate. Novels, games, screenplays, history, journalism, self-defense, and ordinary human frustration all involve violence, and a model that refuses them is broken. A model that helps someone plan real harm is worse. Your job is to tell the difference, case by case, and to explain your reasoning clearly enough that it can train a model.
You will read user requests, model responses, and conversation history, then decide which policy category applies and whether the model's response was appropriate. The interesting cases are the close ones: a torture scene that is either a chapter of a thriller or an interrogation manual with character names; a message that reads as venting about a boss or as a plan; a "realistic" combat question from a novelist that is also a real-world capability question. One word, one contextual detail, or one shift in intent changes the answer.
We are looking for people who already have strong instincts about violence in at least one of these areas: how it works in fiction, how it works in the real world, or how it shows up in people who are struggling. You do not need all three. You need one deep and the judgment to learn the rest.
This is not rote annotation. Policies cannot anticipate every edge case, and good evaluators do not apply them mechanically. You will balance policy text and intent with customer expectations, conversation context, precedent, and team calibration.
Evaluate user requests and AI model responses involving violence, weapons, threats, and dark fiction within the full conversation context
Distinguish fictional, educational, historical, and defensive violence from requests that seek real-world uplift or express real intent to harm
Assess whether a model's response gives meaningful real-world capability, regardless of how the request was framed
Distinguish expressions of anger, frustration, or dark humor from credible threats or crisis indicators
Select the most defensible classification when a case is genuinely ambiguous, and write concise rationales that cite policy language and conversation details
Write and refine adversarial or borderline prompts that probe where a model draws the line
Identify policy gaps, contradictions, and emerging edge cases, and raise them with project leads and policy teams
Participate actively in calibration discussions; challenge interpretations respectfully and update your judgment when stronger reasoning emerges
Apply customer policy consistently without substituting personal beliefs for the policy standard
Maintain accuracy and attention to detail across repeated, feedback-heavy evaluations
You have spent serious time in violent fiction as a writer, game master, game designer, screenwriter, or editor, and you know what good dark fiction looks like and what a story-shaped extraction attempt looks like
You have real-world exposure to violence and its consequences through military, law enforcement, security, EMS, emergency medicine, or similar work, and you can tell movie logic from what actually works
You have worked with people in distress through crisis lines, counseling, threat assessment, domestic violence advocacy, school safety, or trust and safety, and you know the difference between "I could kill him" and a plan
You use AI tools heavily and have opinions about where they refuse too much, help too much, or miss the point
You notice when one word, contextual detail, or change in intent materially affects the answer
You can hold a strong opinion without becoming attached to being right
You explain judgment calls clearly enough that another person can audit your reasoning
You can separate your personal views from the standard a customer has asked you to apply
You remain careful and consistent during repetitive work with difficult material
You communicate clearly and precisely in writing
Strong candidates may come from fiction writing, game design or game mastering, film and TV, military or law enforcement, emergency medicine, crisis counseling, threat assessment, trust and safety, content moderation, journalism, or law. We care more about how you reason than where you learned to reason. A degree, a clearance, and a technical background are not required.
Published or produced work involving violence: novels, short fiction, screenplays, comics, tabletop or video game content, mods, or fan fiction with an audience
Military, law enforcement, corrections, private security, or armed professional experience
Training or professional experience in crisis intervention, threat assessment, forensic or clinical psychology, or violence prevention
Experience with firearms, martial arts, or other weapons disciplines as an instructor, competitor, or professional
Prior work in AI evaluation, red teaming, data annotation, RLHF, trust and safety, or content moderation
Experience evaluating outputs from ChatGPT, Claude, Gemini, or other language models in a professional capacity
Familiarity with calibration sessions, inter-rater agreement, or adjudication workflows
Prior AI evaluation experience is helpful, but it is not required.
This role involves regular and deliberate engagement with graphic material. This is the core of the job, not an occasional part of it. Evaluations will routinely include depictions of violence, weapons, torture, abuse, threats, self-harm, and death, including violence against vulnerable people, along with the emotional distress that often surrounds it.
The work is conducted within structured evaluation frameworks and professional guidelines, with exposure limits, content rotation, and access to mental health support. Candidates must be able to engage with this material carefully, responsibly, and sustainably while maintaining sound judgment and consistent work quality.
Location: Seattle, WA, onsite Monday-Friday
Compensation: $55-$65/hr
Employment classification: W-2
Schedule: 8:00 AM-5:00 PM PT
Weekly commitment: Monday-Friday
Assignment: Ongoing
Planned start date: September 21, 2026
Benefits eligibility: Benefits eligible
United Kingdom