I launched V1 of a new Umbraco Forms package I initially created at the London Umbraco Meetup AI Hackathon a few months ago.
So let me introduce to you Forms Guard.
The premise:
Most bot spam is already handled by honeypots and reCAPTCHA. But, what still gets through is the polite, well-written sales pitch: SEO offers, web design, lead generation. Some are created by people, but more and more these days are written by AI. They’re hard to filter with keyword rules without also catching genuine enquiries, and for many organisations a blocked genuine enquiry costs more than a spam email.
So, Forms Guard doesn’t block anything when the form is submitted. Every entry is saved as normal in Umbraco Forms and the visitor sees the usual thank-you message. In the background, some simple rules run first: honeypot, blocked domains, blocked phrases.
If none of those decide the entry, an AI model is asked a few plain questions about it, such as whether it’s an unsolicited sales pitch or a genuine enquiry.
You can set thresholds in the package which then approve the entry, quarantine it, or leave it for a person to review. Notification emails only go out when an entry is approved, and quarantined entries can be restored.
For compliance and data protection, only the fields you allow are sent to the model, email addresses are reduced to just their domain, and the package’s own tables don’t store any field values, so no personal data gets sent to any external models.
The default model is the new trendy kid in town - Jev from TypeSafe AI - and you can also use Umbraco.AI if you aren’t allowed to use Jev from a security perspective. It needs Umbraco 17 and .NET 10.
I’d love it if some people could try it out and let me know a) if it works and b) is it useful!!!