Email security AI
that judges intent
Our own language model — no open weights, by design, for security
Fraud emails with no links and no attachments: MAIL PRISM AI reads between the lines — the intent behind the words — and makes the call.
At 10,000 emails a day, that is 1.536 billion tokens a month.
At no extra cost.
5,120 tokens per email × 10,000 emails a day × 30 days = 1,536,000,000 tokens
MAIL PRISM AI reads and judges each email in 5,120 tokens. Unlike external AI services, there is no per-token charge — it judges on the CPU inside your gateway, without a GPU.
A language model we built from pre-training up
- Architecture
- Transformer decoder · 24 layers · hidden size 1,024 · 16 attention heads
- Training stack
- RMSNorm · RoPE · SwiGLU · MTP · BF16 · FlashAttention
- Judgment training
- Systematic SFT
Not a shrunk or fine-tuned public model. We designed it ourselves, from architecture to training.
Retoin Software products and key technologies
From email security AI to a secure web engine — built and controlled in-house.
Products
- MAIL PRISM AIEmail security AI that reads between the lines and judgesView
- Secure WebCore EngineUltra-light, security-first pure Java web engineView
- Securein Web SSHA high-security terminal in your browserView
Key technologies
Request materials & product inquiry
Just leave your email address. Our engineers will get back to you quickly.