Kassa AI Transparency Notice
You should always know when you're talking to AI, what it can and can't do, and how to reach a real person instead. This page explains exactly that - for anyone chatting with a Kassa-powered assistant, and for the businesses who run one.
1. You're Always Told It's AI
Every chat window powered by Kassa shows "AI Assistant" in its header - visible the moment it opens, before you've typed a single word. This isn't something a business can turn off or word around; it's built into the product itself, not a configurable setting.
The disclosure sits in the chat header itself, not buried in a privacy policy you'd have to go looking for.
Businesses can customize their welcome message, but the "AI Assistant" header label is fixed, independent of that text.
Shown before any message is sent - satisfying disclosure "at the latest at the time of the first interaction," as required from August 2026 under the EU AI Act.
2. How It Actually Works
The chatbot isn't a generic model reciting canned answers. Each business's assistant is grounded in their own content - the pages on their website, and any FAQs they've explicitly added - so answers reflect that specific business, not a one-size-fits-all script. Generating a reply involves sending your message to a third-party AI provider for processing; we choose reputable providers and don't sell your conversation data.
3. What It Can Get Wrong
We'd rather tell you this plainly than let you find out the hard way:
- Responses aren't guaranteed accurate. The AI generates its best answer based on what it's been given - it can still misunderstand a question or get a detail wrong, the same as any AI system.
- Detecting frustration is a keyword-based heuristic, not true understanding. When a conversation gets flagged for a business's attention because a visitor sounds upset, that's pattern-matching on specific words and phrases - not genuine comprehension. It catches most clear cases and occasionally misses subtler ones.
- It doesn't make consequential decisions about you. The assistant answers questions, captures enquiries, and raises support tickets - it doesn't approve or deny anything that affects your rights, finances, or access to services on its own.
4. Reaching a Real Person
Every Kassa chat has a visible "Need to talk to a real person?" option. Using it doesn't connect you to a live agent inside the chat window - it notifies the business directly, and they follow up with you by email. We're explicit about this distinction on purpose: promising "instant live chat" when the reality is "the business gets an email" would be its own kind of misleading.
If the business doesn't already have your email on file, the assistant will ask for one before completing this request - not to be difficult, but because there's genuinely no way to follow up with you otherwise.
5. If You Run a Kassa Chatbot
As the business, you're responsible for the content you configure - your FAQs, your website's own accuracy, how you follow up on flagged conversations and support tickets. The AI disclosure and escalation mechanisms described above apply to your chatbot automatically; you don't need to configure anything for them to work, and you can't turn them off.
6. Related Policies
This notice covers how our AI works and what it discloses. For how we handle personal data more broadly, see our Privacy Policy; for cookies and similar technologies, see our Cookies Policy; for the terms governing your use of Kassa, see our Terms of Use.
7. Contact
Questions about this policy: contact us, or see our Privacy Policy for how we handle personal data more broadly.