GDPR
Regulation (EU) 2016/679
The problemCall recordings are personal data: minimization, confidentiality and control over transfers.
HelmcodeZero logs by architecture and EU-only inference; on-premise for the strictest requirements.
industries · Contact centers
Every interaction costs the same: almost nothing
Voice and interactions on a flat rate, on EU infrastructure with zero logs. Cost per interaction stops scaling with volume, and calls never pass through an opaque API.
the economics
The economics, by design.
Each cost lever mapped to a platform capability. Sovereignty comes by default.
AI Act (transparency)
Regulation (EU) 2024/1689, Art. 50 · from Aug 2026
The problemUsers must be told when they are interacting with an AI system, and AI-generated content labelled.
HelmcodeOpen, documentable models make the disclosure and the audit trail straightforward, on your own terms.
This page is an informational overview, not legal advice. For your obligations and the risk classification of each system, consult qualified legal counsel. AI Act Guide →
what you have to tell the caller
The disclosure rules for talking to a machine.
A contact operation is the clearest case in the whole AI Act: a person talks to a system and has to be told. Article 50 says what to disclose and by when, and the timing is stricter than most scripts assume.
01
Say it, unless it is obvious
Article 50(1) requires that a person be informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed person in the circumstances. The exception is narrow and it is not about convenience: it covers systems authorised by law to detect, prevent, investigate or prosecute criminal offences.
02
At the first interaction, not later
Paragraph 5 fixes the timing: the information goes in a clear and distinguishable manner at the latest at the time of the first interaction or exposure, and it has to meet accessibility requirements. A disclosure buried in a terms page or read out at the end of the call does not satisfy that.
03
The output gets marked too
Paragraph 2 asks providers to mark synthetic audio, image, video or text in a machine-readable format, detectable as artificially generated or manipulated, and it exempts systems performing standard editing functions. Marking happens where generation happens, which is a reason to control that layer.
EU AI Act · Regulation (EU) 2024/1689 Regulation (EU) 2024/1689 of 13 June 2024, Article 50, paragraphs 1, 2 and 5. Article 50 sits outside the high-risk regime: these transparency duties apply to certain AI systems whatever their risk classification. read the report →
use cases
Your most common use cases.
The cases with the most traction in the sector, each with its own page in detail.
Voice & transcription The full open voice pipeline: Whisper to LLM to Kokoro, at a flat rate and inside your network. Customer support Deflection and response at scale, with the cost per interaction flattened. Summarization Automatic call and ticket summaries into the CRM after every interaction. Semantic search & recommendation Semantic search over the knowledge base so agents and bots find the right answer. Document review & QA Quality assurance and compliance review over interaction transcripts. Conversation QA The whole month scored instead of a sample, so coaching starts from a pattern rather than from twelve calls. Classification & triage Intent and queue decided on arrival, which is where deflection is actually won or lost.
Recommended open models.
A starting point per task type. The full guide maps 80 cases to the open model for each one.
- Whisper large-v3 in Helmcode
- The open standard for multilingual call transcription.
- Kokoro in Helmcode
- Natural voice output with minimal latency, closing the voicebot loop.
- DeepSeek V4 Flash in Helmcode
- Response and summaries at volume, on a flat rate.
- Gemma 4 12B
- Millisecond triage and routing, fine-tunable on your ticket history.
We are distilling and quantizing these open models into small, tightly specialised versions, trained for one task rather than for all of them. A model like that runs on less hardware, answers faster and fits where the big one does not, your own datacenter included. If you have a process with volume and stable criteria, that is the conversation we want to have with you.
// faq
Questions, answered.
What the sector's technical, compliance and business teams ask.
How does the cost work at high call volume?
Pricing is a flat rate per API key, not per token. Cost per interaction stops scaling with volume, which is exactly what breaks the economics of a per-token API once voice multiplies your message count.
Can we run the whole voice pipeline in the open?
Yes: Whisper for speech-to-text, an open LLM for the logic and Kokoro for text-to-speech, on EU infrastructure or on-premise, with no closed API in the loop.
Do call recordings comply with GDPR?
Call recordings are personal data. Prompts and audio are never stored (zero logs) and never train models, processed only on EU infrastructure, and on-premise for the strictest cases.
Does the AI Act require disclosing the bot?
Yes. The AI Act (Regulation (EU) 2024/1689, Art. 50) requires telling users they are interacting with an AI system and labelling AI-generated content, from August 2026. An open, documentable stack makes that disclosure and its audit trail straightforward.
// get started
START BURNING TOKENS
Skip the AI infra work. Deploy your first private inference endpoint today.
Flat rate. EU data. OpenAI API compatible.