Your teams want AI. Your data can't go to labs that log prompts and train on them. LibertAI runs open-weight models behind an OpenAI-compatible API, confidential by design and attested in hardware.
Built for the data you can't leak
Use one piece or all four. The same guarantees hold across the stack: nothing stored, nothing trained on, no profile built on your teams.
The leading open-weight models behind a drop-in, OpenAI-compatible endpoint. Requests are processed and forgotten. Sensitive workloads run in Trusted Execution Environments on Aleph Cloud, with remote attestation your security team can verify.
Explore the APILiberClaw runs always-on agents with persistent memory, powered by LibertAI's private inference. Hand them internal documents, tickets or code: the models never store or train on what they read.
Visit LiberClawGive every team a private assistant at chat.libertai.io, and give developers LibertAI Code, a coding agent wired to a private endpoint. Same daily workflow, without pasting your codebase into someone else's training set.
Try LibertAI Code$ libertai code → backend api.libertai.io → private session · nothing stored ▌
Google, Bing and DuckDuckGo queried in parallel, deduplicated and ranked by consensus. No profiling, no search history tied to you. From $0.002 per query, on the same credits as inference.
Explore SearchSince August 2026, AI Act obligations are enforced and data control is an audit question, not a slide. Our answer is short: processed, never stored, never trained on.
LibertAI is operated by Intelligence Artificielle Générale SAS, in France. You contract with an EU entity and receive an EU invoice.
Open-weight models behind an OpenAI-compatible API. Switching in is one line of config, and so is switching out. Your prompts and code stay portable.
Sensitive workloads run in Trusted Execution Environments on Aleph Cloud. Remote attestation lets your security team check the enclave in hardware instead of reading a policy.
One team, one scoped use case. We help you set it up, you measure the results, then decide.
Your stack already speaks OpenAI. Switch the base URL in one line; we help map models and check your tools still work.
Early-stage teams can apply for up to $10,000 in inference credits, reviewed case by case.
Committed volume gets higher rate limits, direct support and invoicing that fits your procurement. Tailored, on request.
Self-serve stays open: per-token pricing, no KYC · company agreements on request · See public pricing
Pilot, migration, agents or a tailored agreement: it starts with one email to contact@libertai.io.
We answer within 48 hours. Tell us your use case and constraints.