Moonbounce Launches with $12M to Give Organisations Real-Time Control Over AI Behaviour
AI / Machine Learning | 4 min read
Moonbounce, the AI control engine that ensures systems behave exactly as designed at any scale, has launched publicly with $12 million in funding. Lead investors are Amplify Partners and StepStone Group (NASDAQ: STEP), with angel participation from PrimeSet and Josh Leslie, former CEO of Cumulus Networks and Gremlin. Founded in 2024 and based in Oakland, California, Moonbounce was built by engineers with decades of experience at Meta, Apple, and Evernote — and its patented control engine addresses one of the most consequential unsolved problems in enterprise AI deployment: the gap between what organisations intend their AI systems to do, and what those systems actually do in production at scale.
"Most companies know what they want their AI or platform to do. The hard part is making sure it actually does it — every time, without exception. That's the problem Moonbounce solves. We give teams precise control over behaviour at the moment decisions are being made, so they can focus on growth instead of firefighting."
— Brett Levenson, Co-founder and CEO, Moonbounce
The Problem: Traditional Moderation Cannot Scale with Generative AI
As generative AI scales across industries, traditional content moderation approaches — built on retroactive review, rigid static policies, and manual human oversight — cannot keep pace with systems making thousands of decisions per second. The experience of Brett Levenson, Moonbounce's CEO, illustrates the depth of the problem. As former head of Meta's Integrity unit, he saw first-hand how human reviewers were expected to memorise a 40-page policy document — machine-translated into their language — and apply it to each piece of flagged content in roughly 30 seconds, making calls that were only "slightly better than 50% accurate." That experience of policy and enforcement being structurally decoupled led directly to the founding insight of Moonbounce: "policy as code" — converting static policy documents into executable, updatable logic tightly coupled to real-time enforcement. Operational, reputational, and regulatory exposure grows alongside any business that relies on moderation uncertainty rather than systematic, predictable control.
How Moonbounce Works: The AI Control Engine
Moonbounce's patented control engine converts content policies into consistent, predictable AI behaviour in real time. The company has trained its own large language model to read customer policy documents, evaluate content at runtime, and deliver a response in 300 milliseconds or less — taking enforcement action immediately rather than routing content to a queue for delayed human review. Depending on the customer's preferences, that action can mean slowing down distribution of borderline content while it awaits later review, or blocking high-risk content outright at the moment of generation. Teams can develop, test, and deploy content policies in days or weeks rather than months, without extensive custom engineering. Moonbounce also offers the Playground — a sandboxed environment at play.moonbounce.io where teams can write and test policy logic, explore edge cases, and see exactly how rule changes will affect outcomes before deploying to production.
Scale, Customers, and the Investment Thesis
Moonbounce is already deployed in production at scale. The platform has processed 1 trillion+ tokens across a customer base of 250 million monthly active users, evaluating 50 million pieces of content daily. Customers span dating platforms, AI chat applications, AI companion platforms, and generative content sites — including Civitai, Dippy, Channel AI, and Moescape. The platform serves three primary verticals: user-generated content platforms such as dating apps; AI companies building characters or companions; and AI image and video generators. The company is led by Brett Levenson (CEO, former head of Meta's Integrity unit) and Ash Bhardwaj (CTO, former engineering leader at Apple who built large-scale cloud and AI infrastructure). Amplify Partners' investment thesis frames the opportunity as moving AI safety from ethical consideration to mandatory infrastructure — with investor Lenny Pruss describing a world where objective, real-time guardrails become the enabling backbone of every AI-mediated application.
"Content moderation has always been a problem that plagued large online platforms, but now with LLMs at the heart of every application, this challenge is even more daunting. We invested in Moonbounce because we envision a world where objective, real-time guardrails become the enabling backbone of every AI-mediated application."
— Lenny Pruss, General Partner, Amplify Partners
Key Takeaways
- • Moonbounce has launched publicly with $12M in funding (led by Amplify Partners and StepStone Group) — an AI control engine that converts content policies into consistent, predictable AI behaviour in real time, closing the gap between what organisations intend their AI systems to do and what they actually do at scale.
- • The founding insight — "policy as code" — emerged from CEO Brett Levenson's experience as head of Meta's Integrity unit, where human reviewers applied 40-page policy documents in 30 seconds at only slightly better than 50% accuracy. The AI era makes this worse, not better: generative AI systems make thousands of decisions per second, far beyond any retroactive or manual moderation approach.
- • The Moonbounce engine reads policy documents, evaluates content at runtime, and responds in 300ms or less — blocking or routing content before it reaches users, not after. Teams can write, test (via the Playground sandbox), and deploy updated policy logic in days or weeks without custom engineering.
- • Already deployed at production scale: 1T+ tokens processed, 250M+ monthly active users, 50M pieces of content evaluated daily across dating platforms, AI companion and character platforms, and AI image/video generators including Civitai, Dippy, Channel AI, and Moescape.
- • The $12M raise signals the market's recognition that AI safety is transitioning from ethical consideration to mandatory infrastructure — with regulatory frameworks for AI governance crystallising within 12–18 months and enterprises in regulated industries (healthcare, financial services, legal) requiring provable, consistent policy enforcement built into AI behaviour before deployment.