Security your IT team can check

Who can get in

Access is by invitation. Everyone signs in with a personal account and can turn on two-factor authentication. Permissions depend on each user's role, and we can end any session at any time.

Connections from Microsoft Copilot, ChatGPT and Claude

Dantive connects to AI assistants through MCP, the open standard these assistants use for business connectors. Your IT team adds it once, like any other connector.

• Each user connects with their own access, which can be switched off for that person alone. Access keys are only ever stored in scrambled (hashed) form.
• The connection can read EPA regulatory data and nothing else. It cannot see user accounts, passwords or settings, and it cannot change anything. This is enforced inside the database itself, not just in our code.
• Every query is time-limited, so nothing can run away.

A record of what happens

Sign-ins, security changes, administrative actions and every question asked through an AI assistant are logged with time and origin. Older records are deleted automatically after set periods. If you need an extract for your users, ask us.

Your data

Dantive runs on public EPA records, so we do not need your company's data. We keep account details and usage logs. Users choose whether their conversation history is saved, and can delete it, or their whole account, themselves.

AI safeguards

Questions are checked for attempts to manipulate the AI, answers pass through guardrails, and sensitive endpoints are rate-limited. We follow OWASP's guidance for AI applications.

Where it runs

Data is encrypted in transit and at rest. The core database and application servers are in the EU, and Dantive is operated from Spain under the GDPR.

Certifications

Dantive does not hold SOC 2 or ISO 27001 certification today. The infrastructure providers we use have their own. We are happy to fill in your security questionnaire: write to info@dantive.com.