Protection level 1 of 4: 01 — NOT SEALED
Fortron Incognito
Use Claude, ChatGPT and Gemini with your identity removed.
Prepaid credit. The AI company still sees the chat, so it is not sealed.
HIDES WHO ASKEDSee credit
Launching soon
Secure & private AI solutions for businesses and individuals.
Fortron AI works just like ChatGPT or Claude. The difference is underneath: every private chat comes with proof from the chipmakers, AMD and NVIDIA, that the conversation stayed private.
Or read how privacy worksFORTRON ARCADEHI 0000
- SPACE INVADERSHOLD THE LINE
- PONGFIRST TO 7
CLICK HERE TO PLAY
FORTRONREADY
Runs on secure processors from
The problem
People are using AI at work. The data is leaking.
Staff want to be more productive, which makes sense. But most of it happens where the company can’t see.
Concern 1
You don’t know where your data is going
Most AI apps don’t make it clear where your chats end up. DeepSeek, for example, has an excellent model, but its app stores user data in China. Italy blocked it for exactly that reason.2
Concern 2
You can’t check what happens
Personal ChatGPT and Claude accounts can be used to train future models. You can’t see what really happens.
How Fortron AI privacy works
Private. Proven.
Every chat runs in a Trusted Execution Environment (TEE): a locked, encrypted area inside the server’s processors. Nobody outside it can see your chat, including the cloud provider, server admins and Fortron AI.
- 0101_
Locked on your device
Your message is encrypted before it leaves your device.
- 0202_
Private inside the TEE
It’s only opened inside the locked area of the processor.
- 0303_
Proven by the chipmaker
AMD and NVIDIA attest the chat stayed private, and the app checks this first.
Sealed server
SEALED AREA (TEE)
CPU — AMD (sealed)Reads documents, tokenises and searches
GPU — NVIDIA (sealed)Runs the AI model
Fortron AIReadable only in here
- Project Harbour — acquisition
- Offer price: $14.2M
- Close: 30 June
- Cloud providerSees only scrambled data
- Server adminSees only scrambled data
- Fortron AI engineerSees only scrambled data
The chipmakers (AMD and NVIDIA) attest that your chat stayed private, and the Fortron AI app checks that proof before sending anything.
- Model
- Qwen3.8-27B
- Ran in
- Australia
- Kept
- Nothing
- Trained on
- No
- Proven (attested) by
- AMD and NVIDIA
- You
- Fortron AI app checks the proof
- Fortron AI network (your rules, routing)
- Private server (TEE)
- Chipmaker attests
Four levels of private AI.
Choose how much protection your work needs.
- Protection level 2 of 4: 02 — DETAILS MASKED IN THE TEE
Fortron Redact
Names and patient details are hidden before sending, then put back when the answer returns.
Redact does its work inside the TEE, so even Fortron AI never sees the real names or details. Claude and ChatGPT only ever see placeholders.
Reduces the risk of exposing personal and patient data.
See an example
- 01 · YOU WRITEOn your device
Draft a referral for Maria Chen, born 14/03/1981, record MRN 448-219.
- 02 · CLAUDE SEESDe-identified
Draft a referral for PERSON_1, born DATE_1, record ID_1.
- 03 · YOU GET BACKRe-identified by Fortron AI
I am writing to refer Maria Chen (MRN 448-219) for a specialist review…
HIDES WHO + WHATSee add-on - Protection level 3 of 4: 03 — SEALED
Fortron Cloud Lite / Pro
Our own private AI chat on Lite and Pro plans. Every chat runs in a Trusted Execution Environment (TEE), with attestation from the chipmaker (AMD and NVIDIA) proving it stayed private.
- FORTRON LITEQwen3.8-27B · from launch
- FORTRON PRODeepSeek V4.1 · coming soon
PRIVATE + PROVENSee plans - Protection level 4 of 4: 04 — SEALED · ON SITE
Fortron Box
Dedicated AI in your building. No cloud. Nothing leaves the room.
- RUNSFortron Lite (Qwen3.8-27B, 4-bit)
- GPU MEMORY48 GB or 96 GB options
STAYS ON SITESee Fortron Box
Ordered from less protection to more protection.
What you get
The best open AI models, plus the frontier ones when you need them.
Run DeepSeek and Kimi without a single word going to China.
- Sealed
- Not sealed
Fortron Lite
Sealed · From launch- Chat
- Gemma 4
- gpt-oss-120b
- Qwen3.8-27B
- Mistral Small 4
- DOCUMENTS
- Read scans, search your own files
- SPEECH
- Transcription, text to speech
- IMAGES
- Z-Image, Qwen-Image
Fortron Pro
Sealed · Coming soon- DeepSeek V4.1
- GLM-5.3
- MiMo-V2.6-Pro
- Kimi K3
Coming soon.
Frontier
Not sealed · Via Fortron Incognito- Claude
- ChatGPT
- Gemini
Used incognito through Fortron Incognito, the VPN for AI, paid with prepaid credit. Not sealed — and we tell you which is which.
For businesses
Book a demo Built for IT and compliance
Twenty personal accounts can’t tell a manager how staff use AI. One Team account can.
Company rules
Decide what’s kept and which country the servers are in.
Admin console
Add or remove staff and see usage.
Audit proofs for every chat
An attestation for every chat, across the company.
Supervised mode
Optionally log chats to the company, for regulated teams.
Single sign-on
Log in with your company account. One invoice.
Fortron Redact
Hide private details from Claude and ChatGPT (add-on).
On site · Fortron Box
Private AI that lives in your building.
A slim, quiet server with an NVIDIA RTX PRO GPU in a 48 GB or 96 GB option, set up with Fortron AI before it arrives. Plug it in and your team gets Fortron Lite without anything leaving the room.
Fortron Pro needs far more GPU memory than a Box holds, so it is only available in Fortron Cloud.
- GPU
- NVIDIA RTX PRO
- GPU memory
- 48 GB or 96 GB options
- Runs
- Fortron Lite (Qwen3.8-27B, 4-bit)
- Your data
- Never leaves the building
- Managed by
- Fortron AI: updates and monitoring
- Pricing
- Hardware and setup, then monthly
Pricing · coming soon
Be first to access our beta.
Monthly subscriptions. No hardware to buy. Join the waitlist now and we’ll email you the day your plan goes live.
FORTRON LITE
Fast, capable private AI for everyday work.
Runs on Qwen3.8-27B today, upgraded as better open models arrive.
US$30per person / month
For businesses. 5-person minimum.
- Admin console
- Company rules: what’s kept, which country
- Audit proofs for every chat
- Supervised mode
- Single sign-on, one invoice
- Unlimited web search
Web search runs from your own device. Only the results go into your private chat. The private server never touches the internet.
FORTRON PRO
COMING SOONOur most powerful private AI.
Runs on DeepSeek V4.1.
US$50per person / month
For businesses. 5-person minimum.
- Admin console
- Company rules: what’s kept, which country
- Audit proofs for every chat
- Supervised mode
- Single sign-on, one invoice
- Unlimited web search
Web search runs from your own device. Only the results go into your private chat. The private server never touches the internet.
Personal vs team
What Team adds
A workplace isn’t just paying for chat. It’s paying for control.
See all team features| Feature | PERSONAL | TEAM |
|---|---|---|
| Private chatattested by AMD and NVIDIA | Included | Included |
| Web searchruns from your own device | Included | Included |
| Company ruleswhat’s kept, which country | Not included | Included |
| Admin consoleadd or remove staff, see usage | Not included | Included |
| Audit proofsfor every chat | Not included | Included |
| Single sign-onand one invoice | Not included | Included |
Dedicated
Contact us
For regulated workplaces that need their own setup.
- Own private server
- Own rules
- Own country
- Dedicated support officer
ADD-ON · FORTRON INCOGNITO CREDIT
Prepaid credit for Claude, ChatGPT and Gemini, at the AI company’s price plus about 30%. Works with any plan.
ADD-ON · REDACT · +US$20/PERSON/MO
Private details hidden before Claude or ChatGPT sees them, and restored in the answer. Redact does its work inside the TEE, so even Fortron AI never sees the real names or details. Claude and ChatGPT only ever see placeholders. Team plans and Dedicated. Uses your Fortron Incognito credit.
Book a demo · Waitlist
It’s AI, but private — and provably so.
We’re close to launch. Teams can book an early demo; individuals can join the waitlist. Pick a plan and we’ll be in touch as soon as it’s ready.
contact@sensei.dev →