Our mail server supports AI‑powered spam filtering using both OpenAI and Gemini models. These engines analyze message content, headers, and metadata to detect spam, phishing, and other unwanted or risky email automatically — while preserving legitimate mail delivery.
How it works
- Incoming messages are scored by AI models for spam, phishing indicators, and unsafe content. Scores are combined with traditional rules (SPF, DKIM, DMARC, blacklists) to make delivery decisions.
- Models run in a controlled environment; only the data required for classification is sent to the chosen provider. Sensitive content handling and retention follow the configured privacy policy.
Supported providers
- OpenAI — versatile transformer models with strong language understanding for complex patterns.
- Gemini — state-of-the-art multimodal models optimized for high accuracy and low latency.
Benefits
- Higher detection accuracy for modern phishing and targeted spam.
- Reduced false positives through contextual analysis of headers and message body.
- Configurable sensitivity to match your organization’s tolerance for risk vs. false positives.
Privacy & security
We minimize data shared with external providers. You can configure:
- Which provider (OpenAI / Gemini) to use for filtering.
- Whether to redact or omit message bodies and attachments before sending for analysis.
- Retention and logging policies for AI classification results.
Customization & controls
- Tune the spam score threshold that marks messages as quarantine or reject.
- Whitelist and blacklist controls remain available for fine grain overrides.
- Review logs and classification decisions to retrain or refine filters over time.
Getting started
Both OpenAI and Gemini API access can be configured from the Admininstration web interface.
- Choose the provider (OpenAI or Gemini) in the admin settings.
- Set the data handling policy (full body, redacted, or headers-only).
- Adjust thresholds and add any necessary whitelists/blacklists.
- Monitor the quarantine folder and logs during the initial tuning period.
Need help?
Contact us for configuration assistance and best‑practice recommendations.