Security and vulnerability disclosure
Last updated: 18 November 2026 · Approved by: Ben Todd
If you find a security vulnerability in the Arcola AI platform, please tell us privately so we can fix it. Do not open a public issue or publish details before we have resolved it.
1. How to report
Email ben@arcolaai.com. Please include:
- a description of the vulnerability
- the steps to reproduce it
- the affected component, if known
- your assessment of its severity, if known
Our contact details are also published at /.well-known/security.txt.
2. What happens next
- We acknowledge your report within 2 business days.
- We send an initial assessment within 5 business days.
- We agree a resolution timeline with you after the assessment, proportionate to severity.
3. Scope
In scope:
- platform.arcolaai.com, the web application and its programming interface under /api/
- auth.arcolaai.com, the sign-in service
- prompt injection: content that makes the platform's language models ignore their instructions, change how another item is classified, or disclose data they should not
- harmful model outputs: a classification, summary or brief from the platform that is harmful, leaks personal data, or is manipulated by the content it processes
Out of scope:
- third-party services we use (for example our hosting, network and code-hosting providers)
- social engineering
- denial of service, and automated scanning that degrades the service
- brute-force attacks and physical attacks
4. Testing rules
Good-faith security testing of the in-scope services is welcome, on these conditions:
- do not access, modify or take data belonging to other users
- do not degrade the service for others
- report every finding to us before any public disclosure
If you need a test account to assess signed-in features, email us and we will set one up.
5. Safe harbour
We will not take legal action against researchers who:
- act in good faith and follow this policy
- avoid accessing or modifying other users' data
- report findings promptly and do not disclose them publicly before they are resolved
- do not exploit a vulnerability beyond what is needed to demonstrate it