Skip to main content
← sales

SafetyKit Scales Risk Agents to 16 Billion Tokens Daily and Expands San Francisco Team

By Zero G Talent•

SafetyKit is automating content moderation and compliance for online platforms, shifting workload away from human review farms to purpose-built language and vision models. According to an enterprise case study from OpenAI, the 10-person startup grew its daily processing volume from 200 million tokens to more than 16 billion tokens over a six-month stretch while evaluating content with over 95 percent accuracy.

The company operates at the intersection of trust, safety, and operational automation. Rather than employing large teams of outsourced contractors to manually review flagged items, platforms deploy SafetyKit's agents to screen content, analyze transactions, and enforce platform-specific guidelines across text, video, audio, and images.

Its customer roster spans high-volume marketplaces, creator networks, and tech platforms, including Substack, Character.ai, Upwork, Eventbrite, Patreon, and Faire.

Automating Platform Defense Across Modalities

SafetyKit designs specialized agents mapped to individual risk domains, covering fraud, scams, policy disclosures, anti-money laundering, and prohibited content. Each piece of incoming data routes directly to the agent best configured for that category.

For visual enforcement, the company's Scam Detection agent analyzes layouts, embedded phone numbers, and QR codes inside product images using GPT-4.1. For regional requirements, the Policy Disclosure agent checks whether seller listings contain mandatory regulatory disclaimers or wellness product notices, using GPT-5 to verify regional compliance before returning structured output.

In addition to general policy enforcement, SafetyKit partnered with NewsGuard to launch Scaled Misinformation Reviews. The system references NewsGuard's curated database of false narratives, known as Misinformation Fingerprints, allowing platforms to detect and triage misinformation across 133 languages while distinguishing between users advancing a claim and those debunking it.

Under the hood, SafetyKit runs an AWS serverless architecture built on Lambda, Step Functions, and SQS, written in TypeScript. The engineering team implements in-house evaluation harnesses to benchmark new foundation models immediately upon release, integrating hosted frontier models, open-source weights, and specialized vision models. When OpenAI launched o3 and GPT-5, the team deployed the models into production pipelines within days, yielding a 10-point benchmark gain on complex vision tasks.

Leadership Rooted in Stripe and Scaled Infrastructure

SafetyKit was founded in 2023 by David Graunke, Steven Guichard, and Alex Rosenblatt, emerging from Y Combinator's S23 cohort. The company is backed by seed funding from First Round Capital, Ribbit Capital, Liquid 2 Ventures, Y Combinator, and Claire Hughes Johnson.

CEO David Graunke previously led engineering for risk reviews at Stripe. During his tenure, he engineered the policy and workflow engine responsible for scaling Stripe's compliance operations from internal reviewers to thousands of external vendor agents across fraud, credit, and financial crimes.

CTO Steven Guichard brings deep systems engineering experience. He served as the first software engineer and later CEO at Carbic, where he developed ultrasonic flow sensors deployed on offshore oil rigs and industrial pipelines worldwide. He also co-founded Thomas Street, an engineering consultancy servicing clients including Cisco, Roche, and DirecTV. Chief Product Officer Alex Rosenblatt leads product direction, guiding how policy engines integrate directly into enterprise client workflows.

In-Person Expansion in the Mission

SafetyKit works entirely in person from an office in San Francisco's Mission district. With token volume accelerating and enterprise accounts growing, the team is expanding its go-to-market motion alongside its engineering capacity.

The sales pipeline is scaling with two dedicated commercial roles. The Business Development Representative (BDR) focuses on outbound outreach across key accounts and buyer personas, qualifying product-qualified leads and managing enterprise sales research onsite. Working alongside business development, the Account Executive takes ownership of the entire enterprise sales cycle from first touch to close and expansion, working in the office 5 to 6 days per week.

To build out brand presence and distribution channels, SafetyKit is hiring a Growth Lead. This role is tasked with standing up the company's marketing motion from zero to one, covering brand, distribution-led growth, field marketing events, and competitive intelligence.

On the technical side, the startup is recruiting a Full Stack Engineer, Intern Summer 2026 to code and deploy production AI systems into enterprise environments at high velocity.

Open Positions at SafetyKit

Role Location Work Mode Compensation
Business Development Representative (BDR) San Francisco, CA, US Onsite $120,000 – $200,000 / year
Growth Lead San Francisco, CA, US Onsite $120,000 – $200,000 / year
Account Executive San Francisco, CA, US Onsite $150,000 – $200,000 / year
Full Stack Engineer, Intern Summer 2026 San Francisco, CA, US Onsite $12,500 – $16,500 / month

See all open roles at SafetyKit