VoltageGPU, Sovereign Agentic AI for Regulated Industries
Confidential AI agents for legal, finance & compliance teams
VoltageGPU is a sovereign agentic AI platform for regulated industries. Nine specialised agents, Sovereign Legal AI (EU-sovereign Claude-for-Legal alternative), Contract Analyst, Financial Analyst, Compliance Officer, Medical Records Analyst, Due Diligence Analyst, Cybersecurity Analyst, HR Analyst, Tax Analyst, are an AI model following a review checklist, not a lawyer or advisor. Outputs are not legal advice. Inference runs in Intel TDX on a third-party TEE inference network. Our servers see your prompts and documents in clear to relay them; chats are kept in your browser, not in our database.
Why sovereign agentic AI matters in 2026
EU AI Act high-risk obligations apply from 2 August 2026
TEE models, inside Intel TDX at our inference provider
A TEE model from our catalog, fast general agent (Free)
Point CrewAI, LangChain, OpenAI SDK or any MCP client at api.voltagegpu.com/v1. Your agent code runs wherever you want; inference runs in Intel TDX on a third-party TEE inference network. OpenAI-compatible, change one 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.
Compliance
VOLTAGE EI is an EU controller (France, SIREN 943 808 824 00016, Solaize). Native RGPD Article 28 DPA available. Inference runs in Intel TDX; memory is CPU-encrypted inside the enclave. Our servers see each request in clear to relay it; chats are kept in your browser, not in our database; files you add to a matter and scheduled tasks are kept until you delete them. We hold no SOC 2, ISO 27001 or HDS certification, and we do not sign HIPAA BAAs.
Compared to alternatives
VoltageGPU runs only -TEE models, on a third-party TEE inference network, with a $20/mo Plus plan accessible to solo practitioners.
Intel TDX confidential inference · Provider status
Checking provider…
Contacting the TEE inference provider…
What happens to your data?
Standard AI API
✕Processed on shared servers
✕Provider can read your data
✕May be used for training
✕Logs retained for months
✕No cryptographic guarantee
VoltageGPU TDX
Inference runs in Intel TDX
Memory is CPU-encrypted inside the enclave
Never used for training
Chats stay in your browser
Each request runs on one attested TEE instance
How Intel TDX works
1. Our relay
Your prompt and any parsed document reach our servers over TLS. Our servers see them in clear and relay them to our inference provider.
2. Intel TDX enclave
The model runs inside an Intel TDX trust domain on a third-party TEE inference network. Memory is CPU-encrypted inside the enclave, below the OS and hypervisor of that host.
3. Attested instance
Each request runs on one attested TEE instance, which receives the whole prompt. Only -TEE models are offered.
4. Your browser
Chats live in your browser, not on our servers. Files you add to a matter and scheduled tasks are kept until you delete them.
TEE inference network
One attested instance per request
Each request runs on one attested TEE instance.
Your prompt reaches our servers over TLS; they see it in clear and relay it to our inference provider, where the whole request runs on one Intel TDX instance. Memory is CPU-encrypted inside the enclave.
Instances per request
One, attested
Models
-TEE only
Our relay
Sees prompts in clear
Chats
Your browser, 30 days
What we keep on our servers
Since September 8, 2026, chats are not written to our database: they stay in your browser for 30 days after your last message. We keep, until you delete them or your account, the text of files you add to a matter (Beta; the chat does not read matter files yet, and archiving a matter keeps its files) and your scheduled tasks; each scheduled report is emailed to you and not written to our database. The Telegram agent keeps its messages and replies in our database until you delete your account.
Compliance posture
GDPR
Art. 28 DPA, EU controller
Audit trail
Member and config actions logged
CCPA
Chats kept in your browser
–HIPAA
We do not sign BAAs.
–ISO 27001
Not held
–SOC 2
Not held
Verify the infrastructure yourself
Intel publishes the TDX specification and its attestation model. This page does not show a per-request attestation report.