Meta (Llama) and the EU AI Act
Meta releases the Llama family of open-weight general-purpose models under a community licence; the models are typically self-hosted or consumed via clouds and hubs.
What the Act means when you deploy Meta (Llama)
Meta is the provider of the Llama general-purpose models, and the Act's GPAI rules address model providers — with some documentation carve-outs for models released under free and open-source licences (carve-outs that fall away for models posing systemic risk). If you fine-tune or substantially modify Llama and put it on the market, you may take on provider-like duties yourself, so know exactly which weights you run and what you changed.
Questions to send Meta (Llama)
Copy these into your supplier questionnaire. Answers in writing beat answers on a call — your Article 26 file needs evidence, not recollection.
- Q1
Which Llama licence terms apply to our intended commercial use, and what usage restrictions do they impose?
- Q2
What model documentation (model card, training-data summary, limitations) is published for the exact Llama version we deploy?
- Q3
If we fine-tune Llama, what does the licence require regarding naming, redistribution and downstream terms?
- Q4
What is the recommended channel for reporting model vulnerabilities or misuse issues to Meta?
Documents to request for your file
- Instructions for use / product documentation for the exact service and version you deploy
- Model documentation and information for downstream providers, per model in use
- Data processing agreement, sub-processor list and processing locations
- Security documentation: certifications, retention defaults and access controls
- Written statement on whether your inputs and outputs are used for training, and how to opt out
- Licence terms for the exact version you deploy, including commercial-use conditions
Which of these you can actually obtain depends on your contract and tier — treat the list as a checklist to work through with the vendor, not as an entitlement.
Articles to have on the table
- Art. 53GPAI provider duties (the vendor's, not yours)
Meta (Llama) is described here as a provider of general-purpose AI models. Technical documentation, information for downstream providers, a copyright policy and a training-content summary are obligations of the model provider — request the resulting artefacts for your own file.
- Art. 26Deployer obligations
If you use this vendor's technology professionally in a high-risk context, you must follow the instructions for use, assign competent human oversight, keep logs and monitor operation.
- Art. 25When you become the provider
Putting your own name on the system, changing its intended purpose or substantially modifying it moves you from deployer to provider — with Article 16 obligations attached.
- Art. 50Transparency towards people
Where people interact with an AI system, or where content is AI-generated or manipulated, disclosure and — in defined cases — machine-readable marking are required. This duty sits with you, not the vendor.
Most buyers of AI land in the deployer role. Read the full deployer obligations under Article 26 — or the overview of all operator roles if you are not sure which one you hold.
Go to the source
Verify everything on this page with Meta (Llama) directly. Public pages change; profiles do not update themselves.
Important: what this page is not
This profile is based on publicly available information about Meta (Llama) and describes only who they are and what they offer. It makes no assessment of Meta (Llama)’s compliance, contains no rating or certification claim, and is not endorsed by or affiliated with Meta (Llama).
EU AI Act obligations depend on how you deploy this vendor's technology and on your role (provider, deployer, importer, distributor). Verify every fact and commitment directly with the vendor before relying on it — this profile is a starting point, not legal advice.
Nothing here is legal advice. Confirm the facts with the vendor and your own counsel before you rely on them.
Run the free 2-minute checker — deterministic classification, no sign-up.