Skip to content
Visork
Request pilot access

Questions we get on scoping calls

If you can't find what you're looking for, get in touch.

Detection & technology

What does Visork detect right now?

The first production module is image-based CSAM (child sexual abuse material) detection. Detection of violence, deepfakes, and nudity in images, plus grooming detection in text, is on the roadmap and will roll out as separate modules.

How does Visork detect content?

Visork uses a machine-learning model trained to recognise CSAM and other harmful imagery — including new and AI-generated content. Other tools in the space (PhotoDNA, Project Arachnid, Cloudflare's CSAM Scanning Tool) work by matching against databases of known files. The two approaches are complementary and can run side-by-side; Visork covers the case where the content has never been seen before.

How do you handle false positives?

Every analysis returns a probabilistic score, not a binary verdict. You set the threshold per module to fit your platform's risk profile, and the model is calibrated continuously against representative data. False positives are treated as a first-class quality metric — not an afterthought.

Integration

How do I integrate Visork?

After a short scoping call your platform is verified and issued an API key. From there it's one REST endpoint — POST images, JSON in, JSON out. The integration is engineered to fit a backend or moderation pipeline in an afternoon.

How fast is the API?

The system is engineered for sub-200ms end-to-end response, including inference. We track p50, p95 and p99 as first-class SLO metrics, and publish them on the trust portal once the production setup is fully under load.

Do you support video or text?

Today, Visork analyses images. Video, text (chat / comment / grooming detection), and multimodal image+text models are on the roadmap. Webhook notifications are also planned alongside REST.

Compliance & getting started

Where is my data processed?

Inference, storage, and audit logs run inside the EU. Data residency is a design choice, not a paid add-on, and it differentiates Visork from US-based competitors.

Is my content stored or used to train models?

Not by default. Images are analysed entirely in memory (RAM), never written to disk for the analysis itself, and deleted promptly after the verdict. Retention happens only as an explicit step afterwards: content flagged as likely harmful may be held for human review, and material a reviewer confirms is fingerprinted with an irreversible perceptual hash so the same material is caught faster across the service; where its legal qualification justifies it, a finding and its metadata may be preserved for competent authorities. Training on customer content happens only if a platform explicitly enables it — by individual agreement or in the settings. If a free tier launches, its defaults may differ and will be stated at sign-up, with a switch to turn them off.

Do I get audit logs?

Yes. Every analysis is logged with its score, threshold, and module — exportable in a format suited to legal or regulatory review. Compliance reporting is a built-in part of the product, not an afterthought.

Does using Visork make my platform GDPR or DSA compliant?

No — no detection vendor can promise that, and you should be wary of any that does. Visork provides the technical layer: detection scores, verdicts, and an exportable audit trail that document how your platform responds to illegal content. Your policies, moderation decisions, and reporting obligations under the GDPR and the Digital Services Act remain your own — Visork is built to be a well-documented component of that process. For the wider picture, see our guide to the DSA.

Can I try it for free?

Yes. Every pilot opens with 100 free credits — one credit per analysis — so you can measure detection quality and integration effort before any commercial terms. Access is granted after a short scoping call rather than through self-serve signup: a CSAM detection endpoint isn't something we hand to an unverified account. Request pilot access to start one.

Wherever users upload, the same primitives apply

One endpoint, one verdict format, one audit trail. These are the segments where user-generated imagery carries the most legal and regulatory risk under the DSA — and where the integration looks identical.

  • Social & communication

    Stop harmful imagery before it ever surfaces in a feed.

    Score every upload at the edge of your moderation pipeline. Block the obvious, route the unclear, let the rest through.

  • Hosting & cloud

    Screen every upload without leaving the EU.

    A drop-in API for object storage and upload pipelines. Inference, storage, and audit logs are EU-hosted.

  • Dating platforms

    Photo screening at sign-up, frictionless for the user.

    Hold profile photos at the gate. A sub-200ms response target keeps onboarding smooth, even at peak traffic.

  • Communities & forums

    Moderate at scale without growing the team.

    Quarantine harmful uploads automatically and surface only the borderline cases that genuinely need a human.

  • E-commerce & marketplaces

    Catch policy violations the moment listings go live.

    Run listing photos through one endpoint before they're indexed. Fewer takedowns after the fact, less exposure for your review team.

  • Content sharing

    Quarantine quietly, escalate on your own terms.

    Flag high-risk uploads for asynchronous review. Every verdict is recorded with its score and threshold, and exportable for your compliance process.

Request pilot access See how pricing works
100 credits included · EU-hosted · Onboarded selectively