What is VoltageGPU Confidential Chat?

A ChatGPT-style interface whose models run inside Intel TDX on a third-party TEE inference network. Prompts reach our servers over TLS; our servers see them in clear and relay them to the model. Chats are kept in your browser. Memory is CPU-encrypted inside the enclave. We do not train on user data.

Do I need an account to try it?

No. Anonymous visitors get 3 free messages on a TEE model from our catalog without any signup, no credit card required. After the free trial, signup unlocks 50 free messages per month, Plus $20/mo unlocks 2000 messages with Qwen3.5-397B 256K context.

How is it different from ChatGPT Plus?

Same $20/mo price point, but the model runs inside Intel TDX at our inference provider, where memory is CPU-encrypted. Requests pass through our relay, which sees them in clear (TLS ends at our host). Web chats are kept in your browser; files you add to a matter, scheduled tasks and Telegram agent messages are stored on our servers. EU company (France, SIREN 943 808 824), native GDPR Article 28 DPA available.

Which TEE models are available?

3 TEE models, each running inside Intel TDX at our inference provider: a TEE model from our catalog (Free), Qwen3.5-397B-A17B-TEE (256K context, flagship, Plus/Starter/Pro), DeepSeek-V3.2-TEE (163K context, reasoning model, Enterprise).

Can I use it for regulated work (legal, medical, finance)?

Yes. Beyond the generic chat, 9 specialized agents are available: Sovereign Legal AI, Contract Analyst, Financial Analyst, Medical Records, Compliance Officer, Due Diligence, Cybersecurity, HR, Tax. GDPR Article 28 DPA available on request. We hold no SOC 2, ISO 27001 or HDS certification. Agents are an AI model following a review checklist, not a lawyer; outputs are not legal advice.

Yes, api.voltagegpu.com/v1 is OpenAI-compatible for chat completions and billed per token, with live per-model prices on voltagegpu.com/models. Drop-in replacement for the OpenAI chat completions API (change base_url). Requests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain.

Is private fine-tuning available today?

No. Fine-tuning in Intel TDX is not available yet; this page is a waitlist. The planned design runs LoRA training inside an Intel TDX trust domain, where memory is encrypted by the CPU. We will publish the exact data flow before the technical preview opens.

Which base models will I be able to fine-tune?

The preview is planned for LoRA adapters on open-weight base models such as Qwen3 and DeepSeek. The final list will be confirmed when it opens.

Is this RGPD-compliant?

VOLTAGE EI is an EU company (France, SIREN 943 808 824) and offers a native RGPD Article 28 DPA. Compliance depends on your own processing, and fine-tuning itself is not available yet.

Private LLM Fine-Tuning in Intel TDX, Waitlist

LoRA fine-tuning inside Intel TDX is not available yet

VoltageGPU plans to offer LoRA fine-tuning inside Intel TDX, where memory is encrypted by the CPU and the host's hypervisor cannot read enclave memory. It is not available today: this page collects waitlist signups for the technical preview. EU company (VOLTAGE EI, France), native RGPD Article 28 DPA.

Why fine-tune inside a TEE

Train on sensitive records in hardware-isolated memory
Base weights stay open (Qwen3, DeepSeek), which limits vendor lock-in

Planned workflow

Upload a training set over TLS
LoRA training runs on an attested Intel TDX instance
You receive the LoRA adapter weights

Related resources

Private Fine-Tuning in TDX

Waitlist

Waitlist · Fine-tuning inside Intel TDX, not available yet

Fine-tune AI on your
most confidential data.

Planned: your training data is processed inside an Intel TDX enclave, where memory is encrypted by the CPU, and you receive the LoRA adapter weights. Not available yet. Join the waitlist for the technical preview.

For legal, medical, financial, and HR teams that want custom models trained in hardware-isolated memory.

PLANNED ARCHITECTURE, PRIVATE FINE-TUNING IN TDX
STEP 1Your training data
Encrypted in transit (TLS)
STEP 2Intel TDX EnclaveCONFIDENTIAL
Memory encrypted by the CPU · attested Intel TDX instance
STEP 3LoRA training inside enclave
Base model + your data → LoRA adapter weights
STEP 4LoRA adapter weights
Returned to you; delivery format confirmed before the preview opens
Planned: training data not kept after the job
Planned: attested Intel TDX instance
Not available yet, waitlist only

WHY THIS IS DIFFERENT

Hardware-isolated training (planned)

Training is planned to run in Intel TDX, where memory is encrypted by the CPU and the host's hypervisor cannot read enclave memory.

You keep the adapter (planned)

The LoRA adapter weights are returned to you. The delivery format will be confirmed before the preview opens.

Attested hardware (planned)

Jobs are planned to run on an attested Intel TDX instance.

Waitlist only

Fine-tuning is not available today. Nothing on this page runs yet; join the waitlist for the technical preview.

Request early access

Fine-tuning is not open yet. Leave your email and we'll contact you when the technical preview opens.

ESTIMATED PRICING

Starter$299/run

Up to 1K training examples · Qwen3-32B base · 1 LoRA export

Professional$899/run

Up to 10K examples · Qwen3.5-397B base · 3 LoRA exports

EnterpriseCustom

Volume and terms agreed per contract

* Pricing not final. Based on GPU hours of Intel TDX compute.

USE CASES

LegalClassification

Legal document classifier

Fine-tune on your firm's proprietary contracts and case law in hardware-isolated memory.

HealthcareHIPAA

Medical entity extractor

Train a specialized NER model on patient records inside an Intel TDX enclave.

FinanceM&A

Financial risk scorer

Your internal deal data + LoRA training = a proprietary risk model your competitors can't replicate.

HRCompliance

HR policy analyzer

Fine-tune on confidential HR documents to flag potential policy issues.

Want to see it running now?

Try confidential inference in the Playground, available today.

Open Playground

Fine-tune AI on yourmost confidential data.

Fine-tune AI on your
most confidential data.