What is VoltageGPU Confidential Chat?

A ChatGPT-style interface running inside Intel TDX hardware enclaves. Prompts and responses are encrypted in CPU memory at runtime and never leave the sealed Trust Domain. VoltageGPU cannot read user data, cannot log it, cannot train on it.

Do I need an account to try it?

No. Anonymous visitors get 3 free messages on Qwen3-32B-TEE 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 Qwen3-235B 262K context.

How is it different from ChatGPT Plus?

Same $20/mo price point but sealed in Intel TDX hardware. OpenAI trains on user data by default. VoltageGPU Plus cannot technically read user data — memory is encrypted with CPU-fused keys and even the hypervisor is outside the trust boundary. EU company (France, SIREN 943 808 824), native GDPR Article 28 DPA available.

Which TEE models are available?

3 models sealed in Intel TDX: Qwen3-32B-TEE (40K context, fast), Qwen3-235B-A22B-Instruct-2507-TEE (262K context, flagship, Plus/Pro), DeepSeek-R1-0528-TEE (163K context, reasoning model, Enterprise).

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

Yes. Beyond the generic chat, 8 specialized agents are available: Contract Analyst, Financial Analyst, Medical Records, Compliance Officer, Due Diligence, Cybersecurity, HR, Tax. GDPR Article 28, HIPAA, DORA, NIS2 ready. DPA available on request.

Yes — api.voltagegpu.com/v1 is OpenAI-compatible, pay-per-token starting at $0.15/M input tokens on Qwen3-32B-TEE. Drop-in replacement for the OpenAI SDK (change base_url).

How is hallucination prevented?

Every citation produced by the agent is resolved against primary jurisprudence sources before reaching the user. Citations that cannot be resolved against Légifrance, EUR-Lex or court bulletins are stripped automatically. The agent does not eliminate hallucination risk entirely — no LLM does — but the guard makes its scope explicit and its failures recoverable.

How does this compare to Lexis+ AI or Doctrine?

Lexis+ AI runs on Microsoft Azure (US infrastructure). Doctrine is a French jurisprudence database with its own AI features but no hardware sealing. VoltageGPU runs the same primary-source citation guard inside an Intel TDX enclave we operate in the EU, with attestation evidence per session.

Which corpora does the agent cover?

EU primary law (treaties, regulations, directives, EUR-Lex jurisprudence), French primary law (Code Civil, Code de Commerce, Code du Travail, Code de la Consommation, etc.), Cour de Cassation and Conseil d'État decisions via Légifrance, and curated CJEU case law. EU member-state expansion is on the 2026 roadmap.

EU + FR jurisprudence · Citation-guarded

Legal Research AI —
Sovereign LLM for EU Jurisprudence.

Memo, brief and pleading templates grounded in EUR-Lex and Légifrance. Citations resolved against primary sources before they reach you.

Available on Pro $1,199 and Enterprise plans.

See plans All legal agents

Jurisprudence corpus

Primary sources, refreshed regularly.

The agent is grounded in EU and French primary sources only. No mid-tier blogs, no unverified summaries. Citation extraction runs on EUR-Lex, Légifrance, and curated CJEU and Conseil d'État decisions.

EU primary law

· TFEU + TEU consolidated
· Regulations (current)
· Directives (transposed FR)
· Charter of Fundamental Rights

EU jurisprudence

· CJEU decisions
· General Court rulings
· Advocate General opinions
· Pending references for preliminary ruling

French codes

· Civil
· Commerce
· Travail
· Consommation
· Construction
· Pénal
· CGCT

French case law

· Cour de Cassation
· Conseil d'État
· Cours d'Appel selected
· Conseil Constitutionnel

Regulatory authorities

· CNIL délibérations
· AMF doctrine
· Autorité de la concurrence
· ARCEP / ARCOM

Soft law

· CCBE guidance
· CNB position papers
· EDPB guidelines
· EU Commission communications

Hallucination guard

Citations resolved before they reach you.

Every output goes through a deterministic citation-check pass. The agent enumerates all cited references — court decisions, code articles, regulations, AG opinions — and resolves each one against the primary-source corpus. References that cannot be matched are stripped automatically. The remaining citations are guaranteed to point to a real, verifiable source.

Step 1 — Enumerate

Extract every citation pattern in the draft response.

Step 2 — Resolve

Match each citation against Légifrance, EUR-Lex, court bulletins.

Step 3 — Strip

Remove unmatched references before the response is returned.

Step 4 — Stamp

Each retained citation is labelled with publication date and source URL.

Output templates

Memo, brief, pleading, client letter.

Legal memo (IRAC)

Question presented · short answer · issue / rule / analysis / conclusion. Citations verified against primary sources only.

Pleading skeleton

Assignation or conclusions structure with factual background, legal basis (codes + jurisprudence), prayer for relief.

Client letter

Plain-French explanation with legal basis broken out separately. Cites stripped from main body, listed in annex.

Comparative analysis

EN vs FR vs DE on a focused question (e.g. non-compete enforceability, indemnity caps, IP assignment).

Comparison

VoltageGPU vs Lexis+ AI vs Doctrine.

FEATURE	VOLTAGEGPU	LEXIS+ AI	DOCTRINE
Hosting	Intel TDX, EU	Microsoft Azure (US)	EU cloud (no TEE)
Hallucination guard	Primary-source citation check	Closed corpus only	Closed corpus only
Coverage	EU + FR primary sources	Multi-jurisdictional	FR jurisprudence
Hardware sealing	Yes (TDX, attested)	No	No
API	OpenAI-compatible	Limited	Yes
Solo/small-firm pricing	$1,199 Pro	Enterprise only	From €99/seat

Pricing

Pro

$1,199

Up to 10 seats · OpenAI-compatible API

Enterprise

$3,499

R1-TEE reasoning · SSO · attestation receipts

FAQ

Frequently asked questions.

Has there been a real legal sanction for AI hallucination?

Yes — Mata v. Avianca (US, 2023) is the textbook case. A New York attorney filed a brief that cited six fictitious decisions hallucinated by ChatGPT. The court issued sanctions, mandatory CLE on AI use, and the case became required reading in legal-ethics curricula. Subsequent cases in Colorado, Illinois and Australia have followed similar patterns. The technical answer is to constrain the agent to verifiable primary sources and reject anything else — which is what the citation guard does.

Can I trust the agent on niche French law areas?

The agent is strongest on commercial, employment, data protection and IP — areas with heavy primary-source coverage on Légifrance and well-codified jurisprudence. It is weaker on highly local administrative law, recent fiscal positions and family law, where regional variation and rapid evolution exceed the corpus refresh cadence. Always verify outputs against a canonical source before relying on them.

Does the agent train on my queries?

No. Inference happens inside an Intel TDX enclave with zero retention by default. Queries do not enter any training pipeline. There is no fine-tuning on user data — the underlying Qwen3-235B-TEE base weights are frozen and shared across all customers, and the citation-guard pass is deterministic, not learned.

How current is the corpus?

Primary law and codified texts are refreshed weekly from Légifrance and EUR-Lex. Jurisprudence is refreshed bi-weekly with a four-week lag for selected cases pending consolidation. The agent labels every citation with its publication date so the lawyer reviewing the output knows the recency.

Verifiable citations or none at all.

Try the Legal Research agent on a Pro plan. Hallucination guard, primary-source corpus, EU-controlled.

See Pro plan Verify the enclave

Legal Research AI — Sovereign LLM for EU Jurisprudence

Hallucination-guarded research across French and EU case law

Corpora

Hallucination guard

Templates

Pricing

Related resources

Legal Research AI —
Sovereign LLM for EU Jurisprudence.

Primary sources, refreshed regularly.

Citations resolved before they reach you.

Memo, brief, pleading, client letter.

Legal memo (IRAC)

Pleading skeleton

Client letter

Comparative analysis

VoltageGPU vs Lexis+ AI vs Doctrine.

Frequently asked questions.

Verifiable citations or none at all.

Legal Research AI —Sovereign LLM for EU Jurisprudence.

Primary sources, refreshed regularly.

Citations resolved before they reach you.

Memo, brief, pleading, client letter.

Legal memo (IRAC)

Pleading skeleton

Client letter

Comparative analysis

VoltageGPU vs Lexis+ AI vs Doctrine.

Frequently asked questions.

Verifiable citations or none at all.

Legal Research AI —
Sovereign LLM for EU Jurisprudence.