Amid the rapid technological revolution, digital content moderation has become one of the most complex challenges facing social media platforms and technology companies worldwide. Relying solely on human elements is no longer sufficient to control the massive influx of data and constantly renewing content around the clock, creating an urgent need to innovate radical solutions that ensure user safety and provide a healthy digital environment free from ongoing risks and threats. From here emerged the idea of establishing specialized companies that use advanced technologies to provide innovative and proactive solutions that bypass slow traditional methods and keep pace with the growth rate of artificial intelligence.
From the halls of Apple and Facebook to pioneering proactive solutions
When Brett Levinson left Apple in 2019 to take charge of business integrity at Facebook, the social media giant was at the epicenter of the fallout from the infamous Cambridge Analytica scandal that shattered user trust. At the time, Levinson believed he could simply fix the platform’s content moderation problem using better, more sophisticated technology. But the problem, as he quickly realized through his fieldwork, was much deeper than fleeting technical challenges.
In traditional systems, human reviewers were asked to memorize a complex 40-page policy document machine-translated to fit their local languages. Afterward, reviewers were given only about 30 seconds for each piece of flagged content to make a critical decision—not just whether the content truly violated the rules, but what strict action to take: whether to ban the content, permanently ban the user, or limit the post’s reach. According to Levinson, those rushed and pressurized decisions were “only slightly better than a 50 percent accuracy rate,” representing a clear shortcoming in protecting digital communities.
The rise of Moonbounce and new funding
This type of reactive and delayed approach is no longer sustainable in a world full of smart and well-funded hostile actors. The rise of AI-powered chatbots has exacerbated the problem unprecedentedly, leading to a series of high-profile incidents that made headlines and stirred public opinion.
Levinson’s frustration with these sterile mechanisms led to the invention of the concept of “policy as code,” a pioneering method aimed at transforming static, written policy documents into dynamic, actionable, and updatable programmatic logic closely tied to actual enforcement operations. This profound vision led to the founding of Moonbounce, which recently announced it has raised a massive $12 million in funding. The funding round was led by Amplify Partners and StepStone Group, reflecting investor confidence in this unique technical solution.
How it works and target sectors
Moonbounce works alongside other companies to provide an additional, robust layer of security wherever content is generated, whether that content is created by a human user or generated by artificial intelligence. The company has trained its own large language models to accurately examine client policy documents, evaluate content at actual runtime, and then deliver a decisive response within 300 milliseconds or less, taking necessary action immediately without any noticeable delay.
Today, the company serves three main technology sectors: platforms dealing with user-generated content such as dating apps, AI companies building virtual personas or companions, and generative-powered image and video generators. Levinson noted that the platform currently supports over 40 million daily review operations and serves more than 100 million daily active users, with prominent startup clients including Channel AI, Civitai, and interactive roleplay platforms like Depii AI and Moyscape.
Security as a core feature, not just an obligation
Levinson emphasized in his statements that the concept of security “can actually be a strategic product feature itself.” He added: “That wasn’t the case before because it was always considered an afterthought added at the end, rather than something fundamental you could actually build into the core of your tech product from the early development stages.” Today, AI companies are increasingly looking outside their walls for help strengthening their own security infrastructure. Levinson explained his company’s role: “We represent a third party sitting between the user and the chatbot, so our system isn’t flooded with the intense conversational context like the chat interface itself, allowing us to intervene accurately and effectively.”
Future vision: Iterative steering for user protection
Levinson manages this 12-person exceptional company in close collaboration with his former Apple colleague Ash Bharadwaj. Their next focus is a new and innovative technical feature called “iterative steering.” This sensitive feature was developed in response to tragic and painful cases, such as the 2024 suicide of a 14-year-old boy in Florida after becoming obsessed with talking to a Character.AI chatbot.
Instead of relying on outright, blunt refusals that can be frustrating or provocative when harmful topics arise in conversation, the Moonbounce system intelligently steps in to intercept the conversational path, seamlessly redirecting it and adjusting inputs in real-time to push the chatbot toward providing a supportive and positive response proactively and thoughtfully.
This radical shift in the philosophy of handling digital content marks a watershed moment in the development of safe technology. By consciously moving from mere reactive delay to proactive, intelligent intervention, Moonbounce is establishing a new and advanced industry standard where digital security is not just a heavy regulatory burden, but a core pillar that enhances user trust and supports the growth of technological innovation in a safe, risk-free environment.
Frequently Asked Questions
What is Moonbounce and how was it founded?
It is a startup specializing in providing AI-powered content moderation solutions, founded by Brett Levinson after his extensive experience at Apple and Facebook, aiming to turn traditional safety policies into code ready for immediate execution.
How much funding did the company recently raise and who led it?
The company raised $12 million in an investment round led by Amplify Partners and StepStone Group to support its innovations in digital security.
How does the company’s technology work in content auditing?
The company uses specially trained large language models to read and understand corporate policy documents, then evaluates content—whether text or image—in a fraction of a second (under 300 milliseconds) and makes the appropriate decision instantly.
What is meant by the “iterative steering” technology developed by the company?
It is an intelligent and innovative technology aimed at intercepting harmful or dangerous conversations with chatbots, proactively redirecting them to provide supportive and positive responses to the user, rather than simply rejecting the content outright.
