We're excited to announce SafetyKit’s $27M in fundraising led by Ribbit Capital, with participation from First Round Capital, Y Combinator and angels like Claire Hughes Johnson (former COO, Stripe) and Jon Zieger (former GC, Stripe). Today big tech and finance companies have to rely on armies of low-paid outsourced humans to protect their platforms from fraud, harassment, and crime. It’s super important work, but the jobs are difficult and sometimes even traumatizing — and it’s impossible for companies to keep up. If any human work should be replaced by AI, it’s this. That’s why we’re building SafetyKit. SafetyKit builds AI agents that check for every bad thing you might worry about, plus the things you don’t know about like new scams and sanctions. That means companies don’t need to rely on overworked humans to do their dirty work — they can let AI agents do it instead. Huge companies already rely on SafetyKit in production — companies like Patreon, Eventbrite, Upwork, Substack, and Character.ai. They’re offloading work from humans and solving problems that were intractable without AI. We’re growing our engineering team in San Francisco. It’s the opportunity to do the best work of your life: building real-world AI at scale for some of the world’s biggest companies. Learn more here: safetykit.com/jobs Read more: safetykit.com/resources/safe…

SafetyKit partnered with @OpenAI to evaluate reasoning models with reinforcement learning for SafetyKit's toughest risk and compliance workflows. Traditional LLMs and human agents both fall short here. We saw great results, and we're rolling out out fine-tuned reasoning agents to our customers now. Early access to SOTA models and techniques makes a huge difference for SafetyKit customers, and we're excited to continue partnering with OpenAI!

We worked with the @OpenAI  team to test and evaluate fine-tuned reasoning models in SafetyKit and saw great results, especially with complex instruction following and large context windows. SafetyKit customers need lots of expert-level decision-making across lots of complex policy areas. OpenAI’s reinforcement fine-tuning lets us replace a bunch of nodes in our most complex workflows with a single o3-mini reasoning agent. That simplification lets us expand across policy areas much more quickly, and that’s what our customers need most.