TrollWall AI Scale Human-Centered Moderation Technology with EUR 800K Seed Funding

• TrollWall AI attracted a EUR 800K seed round led by Seed Starter Slovenská Sporiteľňa
• The founding team was inspired by real-world moderation challenges in political and humanitarian communications
• The platform uses localized AI models and human experts for bias-controlled moderation systems
• The ambition is to expand toward AI agent infrastructure for global community management platforms.

This April, the Slovakian automated moderation platform TrollWall AI raised its Seed round of EUR 800K. Slovenská Sporiteľňa’s Seed Starter led the round, joined by Slovak Investment Holding (SIH), and angel investor Miroslav Fedor of HERNWOOD.

From Real-World Moderation Pain to Startup Idea

Tomáš Halász (CEO) realized the need for a comprehensive local automated moderation solution like TrollWall AI in 2022 when he worked as a social media manager for the Vice President of the European Parliament and co-founded KtoPomozeUkrajine, an initiative helping Ukrainians fleeing the war. One of his main responsibilities there was handling messages and comments on social media. As Mr Halász recollects, many were supportive, but a large part was filled with hate, profanity, and toxicity. Together with a team of volunteers, they had to remove those comments manually.

‘I eventually got fed up with it. Not emotionally, but practically. It was inefficient, exhausting and clearly not scalable. When I looked for existing solutions on the market, I couldn’t find one supporting Slovak and other under‑resourced languages, while also not being a black box, without transparency or control for social media managers,’ Mr Halász recalls.

That’s when he was convinced to build our own solution. He reached out to his friends with whom he had already worked on complex IT projects:

• Filip Strýčko (CTO) brought experience from leading large‑scale digitalization, cloud and data projects at EY and Accenture;
• Pavol Bujňák (COO) contributed more than a decade of experience in building and delivering enterprise‑grade products and infrastructure systems for multinational clients;
• Dr Tomáš Koctúr, a seasoned data scientist with a PhD background and years of hands‑on NLP experience, designed and led the AI detection layer.

Together, they built the first version of TrollWall AI for internal use, but soon enough, it got adopted by other social media managers from news organisations, NGOs, and public institutions facing the same problem.

Early Traction & First Institutional Adoption

In March 2023, the founding team incorporated the company while still keeping their jobs and developing TrollWall AI in a bootstrapped mode. Early on, they faced a surprising level of demand: within the first months, the startup got to sign high‑profile users including heads of state, telecom operators, major brands and news media, while also supporting NGOs.

This traction helped the startup raise its first investment of EUR 500K from CB ESPRI Impact One in December 2023. This funding allowed the team to focus on the company fully in 2024. Within half a year, TrollWall AI hired more talent, added support for five more languages with human‑level accuracy, and signed clients in eight European countries.

AI Agents and Multilingual Moderation

By 2026, TrollWall AI has evolved far beyond a moderation tool. Today, it is a full set of AI agents for community management that analyse sentiment, recommend actions and draft replies, while still reliably moderating toxic language that is harmful both to society and to brands. The platform supports 14 languages, including three variants of Spanish tailored for Latin American markets.

TrollWall AI is used by major sports clubs, heads of state, political parties and brands, including former Slovak president Zuzana Čaputová, Czech president Petr Pavel, football club AC Sparta Praha, telecom operator O2 Slovakia, and many more clients across the EU and the US.

‘Our core hate detection is fully developed in‑house by the founders, which gives us full control over the data and the quality of the models. Thanks to this approach, we outperform many competitors and also generic GPT‑based models in moderation quality, not only in Slavic languages, but also in English and German. The biggest remaining bottleneck is understanding local context. Language is deeply local, and slang, irony and cultural references change from country to country. This is why we work with external linguistic agencies and native speakers to train and refine our models,’ Mr Halász tells ITKeyMedia.

Instead of relying on universal translations for dozens of languages, a solution for each language gets trained separately by people who live in a particular linguistic environment and understand expressions that never appear in dictionaries. While it’s significantly slower and more resource‑intensive, the TrollWall team believes that it is the only way to achieve high‑precision moderation in sensitive environments.

‘We audit bias directly in how our AI is trained. For every language, we work with three to five local linguistic experts who are native speakers and live in the target country. Every comment used for training must be labelled by all experts in agreement. If there is no consensus, the comment is excluded from training,’ Mr Halász explains.

This is essential for politically or culturally sensitive content, where meaning depends on local context. According to the team, their clients see how TrollWall AI behaves in real‑world conditions every day and across many cases, they confirm that the platform is trained to moderate hate, not opinions. Legitimate criticism, political disagreement or sensitive discussion is not the target.

Moderation VS Censorship

For TrollWall AI, the line between moderation and censorship is apparent: Moderation removes behaviour that harms others or breaks the discussion; censorship suppresses opinions.

‘Both our experience and existing research show that unmoderated discussions tend to be hijacked by a small but very loud group of toxic users. In those environments, women and people from minorities are far less likely to participate. Ask yourself a simple question: would you join a discussion where every second comment is profanity or hate? Most people would not. They leave, even if they have something valuable to say,’ Mr Halász clarifies.

Consequently, moderation, when done correctly, actually supports freedom of speech. It creates a safe space where more people feel confident to express their opinions. TrollWall AI is trained to remove hate and abuse, not disagreement or criticism.

The team admits that they sometimes have to push back against a client’s definition of harmful content, often early during onboarding. They make a point of explaining what TrollWall AI is for and what it is not: moderating hate and abuse, not opinions.

In some cases, clients question why the AI did not moderate a comment they personally considered toxic. In response, the TrollWall AI team takes a practical approach and shows them what the platform has already moderated for their profiles, often the worst cases of threats, hate speech or coordinated abuse. Seeing the contrast usually makes the distinction clear and clients understand that the system is deliberately conservative with opinions and strict where real harm begins.

Big Tech Moderation Critique

Mr Halász is convinced that moderation on major platforms like Meta or TikTok has become less reliable over time, and our demand has grown year on year. According to him, such platforms always optimize moderation for their own needs, scale and risk management, not for the needs of an individual brand, newsroom or institution. TrollWall AI’s clients, in turn, care about precision, local context and control because the consequences of mistakes are very concrete for them, and the startup builds specifically for those teams. 

TrollWall AI’s approach to training and auditing was recognised with a Trustworthy AI award, highlighting its focus on transparency, local expertise, and human oversight.

Global Expansion and Shift Toward AI Infrastructure 

Seed Starter Slovenská Sporiteľňa’s executive manager Michal Vanovčan, his fund invests in startups that solve real-world problems and have the potential to scale. TrollWall AI meets both of these criteria perfectly—it helps manage the toxic online content that virtually all brands and institutions communicating online face today, and it does it in a way that is applicable across countries and cultures.

‘Our investment in TrollWall AI serves as further proof that systematic startup support in Slovakia works—when it combines capital, mentoring, education, and access to a real market. Our ambition is to build a long-term startup ecosystem and be a partner for teams from their initial ideas through their growth and subsequent international expansion. The story of TrollWall AI shows that when a startup has a high-quality product, an openness to feedback, and the support of the right people at the right time, it can grow from Slovakia to the world stage,’ Mr Vanovčan states.

In view of the new funding, TrollWall AI’s goal is to move from being mainly a CEE player to becoming a global one. At the same time, the platform is moving from being mostly a moderation tool to a full set of AI agents that help clients understand their communities better, engage with followers more effectively and convert attention into real value. In other words, the investor support allows the startup to scale what already works, expand into markets where demand is already coming from, and continue building AI tools that go beyond removing toxic content.

‘We are still improving as a vertical SaaS, but we are also moving towards infrastructure for an agent‑based economy. Our clients deal with hundreds of thousands to millions of comments and messages per month, and we already operate the infrastructure and specialised agents optimised for this scale. In the age of tools like Claude and OpenClaw, we want TrollWall AI to become a go‑to platform not only for human community managers, but also for AI agents acting on their behalf,’ Mr Halász concludes.

As such, moderation is just the entry point for TrollWall AI. If a time comes when discussions become healthier, the platform’s value will shift to understanding sentiment, spotting patterns, prioritising meaningful interactions and engaging the right people at the right moment, and that is where TrollWall AI targets to evolve. In a better online world, the need for good community management does not disappear. It becomes even more important.

As online discourse becomes increasingly difficult to manage, TrollWall AI is positioning itself at the intersection of AI infrastructure, digital trust, and community safety. Its recent EUR 800K Seed round validates the growing demand for precise, locally informed moderation tools and signals investor confidence in a future where brands, institutions, and AI agents alike will require scalable systems for managing online engagement responsibly. By combining human oversight with language-specific AI models, TrollWall AI is building toward a broader vision: healthier digital communities powered by transparent and trustworthy automation.