What is Varselio?
Varselio is a Swedish security service that checks what your website chatbot or AI agent is about to say or do, in real time, and stops prompt injection, jailbreaks and data leaks before they cause harm, asking a human before risky actions.
What it does
- Checks before the AI acts. Your chatbot or agent asks Varselio before it replies or acts. Each message or action gets an answer in real time: go ahead, wait for a human, or stop.
- Stops the dangerous. Manipulated instructions, data leaks and destructive actions are refused on the spot.
- Asks a human. Payments, deploys and anything unfamiliar wait for your yes or no, right on your phone.
- Shows what happens. Live events and alerts in a dashboard, and a notification when something needs you.
- Sets traps for attackers. Fake keys, logins and links are placed where an intruder or a rogue agent would look. If anyone uses one, you get an alert with the evidence.
- Keeps proof. Every decision is kept in a tamper-evident log you can show to customers, auditors or your board.
Who it is for
Companies of any size that run an AI chatbot on their website, a customer-service bot, or AI agents that read e-mail, search data or press buttons on their behalf. It is open to everyone and free to start, so it also suits small businesses without a security team.
Facts
- Currently a free public beta.
- Built and run by Dennis Erichsen in Hallsberg, Sweden.
- Data hosted in the EU. No tracking or advertising cookies. Your data is never sold or used to train AI models.
- Available in English and Swedish.
- Free to start; prepaid passes from 279 SEK (see pricing).
About prompt injection (OWASP LLM01)
Prompt injection is number one on the OWASP Top 10 for LLM applications: LLM01:2025 Prompt Injection. OWASP notes that it is unclear whether fool-proof prevention exists, and recommends layered measures such as input and output filtering, least privilege and human approval for high-risk actions.
Varselio reduces risk, but no security product is 100% secure: it cannot guarantee that every harmful message or action is detected.