Your team keeps their own ChatGPT and Claude accounts. Compliance gets the audit trail. Built for regulated Nordic firms.
Runs in the browser · Metadata only · Hosted in Stockholm
Claude
Gemini
Perplexity
EU data residency · Stockholm
GDPR-first architecture
DORA & NIS2 evidence exports
100% on-device detection
AI use in most firms started bottom-up and stayed there. A third of workplace ChatGPT sessions run on private logins, outside SSO, outside retention policies, outside anyone's view.
Gartner expects at least 80% of unauthorized AI transactions through 2026 to come from employees breaking internal policy to work faster, rather than from attacks. In a regulated firm, a personnummer pasted to save 10 minutes is a GDPR incident with a client's name on it.
When a client, your auditor or Finansinspektionen asks how AI is used, a document describes the rules. It has no record of a single prompt.
Vaktor closes the gap where it opens: in the browser, before anything leaves the device.
The field above is editable. Masking runs in your browser, which you can confirm in the network tab (F12). The demo recognizes Swedish personnummer, org numbers, IBANs, emails and a sample client list. Demo data uses official Swedish test numbers.
A browser extension on each laptop. A team is live in an afternoon, working on chatgpt.com, claude.ai and gemini.google.com as before.
Personnummer, org numbers, IBANs and your client list are replaced before a prompt leaves the device. The same entity gets the same placeholder every time, so the AI can reason about the case and the answer stays useful.
Compliance gets a metadata audit trail: categories, counts, timestamps. Export it when Finansinspektionen, your auditor or a client asks how AI is used.
Asset managers, law firms and fintechs with 30 to 150 people, where one leaked client name can end the relationship. Built for the CEO or CTO who owns the risk: your teams use AI today, and Vaktor lets you say yes on your terms.
Vaktor sits outside the prompt path by design. Detection runs on the device, and our backend receives metadata only. Your security consultant can confirm it with standard network monitoring. The claim is checkable, and that is the point.
The browser is where AI work happens today, so that's where Vaktor starts. The same control point extends to desktop assistants, coding tools and agents acting on your firm's behalf. Govern access, mask what's sensitive, prove what happened. One layer, every model.
The models know this problem well. Put the question to the tool itself and see what comes back.
Install across 10 to 25 seats and load your client list. Your team signs into their usual AI accounts and keeps working.
Masking runs live. We tune detection against your real patterns, with a short check-in each week.
You get the audit trail export and a joint review: what was masked, what it means for your AI policy, and what a full rollout looks like.
Fixed fee, credited toward the annual plan if you continue.
30 minutes. We'll show you what would have stayed on the laptop, and what a four-week pilot looks like for your firm.
or email directly: [email protected]