We're building the safety layer for the internet.
We protect platforms from the content they can't afford to host.
Our mission
Visork develops scalable, accurate AI for real-time detection of illegal content. The first production module is CSAM detection in user-generated images — because that's where the legal, reputational, and human cost of inaction is highest. The mission is to set the European reference point for AI-driven content safety.
We focus on performance, reliability, and integration that takes hours, not quarters. Our tools sit inside the upload pipelines of social platforms, hosting providers, dating apps, marketplaces, and forums — anywhere user uploads can hurt someone.
Our vision
The internet's centre of gravity is moving online — every interaction, every transaction, every conversation. So is the harm. Our vision is to be the European reference point for online safety: the platform that defines what good looks like for AI-driven detection and protection.
EU-first isn't a positioning angle. It's a structural choice — about where data lives, who oversees it, and which legal framework you're answering to when something goes wrong.
Four values, in order of weight.
Responsibility
AI is reshaping the internet faster than any oversight can keep up. We treat that as a duty, not a tailwind. Every model decision we ship has a person willing to defend it.
Safety
Everything we build exists to protect people — children first. That's the lens for product trade-offs, not a marketing line we attach after the fact.
Transparency
Probabilistic scores, audit logs, exportable reports. If a verdict has consequences for someone, the customer should be able to explain why — and so should we.
Innovation
The shape of online harm keeps changing — new content, new patterns, new ways to evade detection. Innovation means investing in models that adapt as the problem does, not as it looked a year ago.
The team
Visork is founded and led by Filip Šedivý out of Prague. The project began in 2021 and is shaped by direct collaboration with law enforcement and the people who moderate user-generated content for a living.
Solo-founder for now — by choice, not by accident. Technical, legal, and commercial decisions sit at one table, which means we move quickly and take a clear position on the things that matter: data residency, model accountability, and the ethics of what we choose to detect.
We'll grow the team as customers come on board. If trust & safety, applied ML, or compliance engineering is where you want to spend the next decade, we'd like to hear from you.
- Block illegal imagery before it reaches your users
- Drop-in REST API — integration in hours, not quarters
- EU-hosted end to end — GDPR compliance built in, not bolted on
- Auditable scores you can defend to regulators or in court