AWS (Bedrock) and the EU AI Act
Amazon Bedrock is AWS's managed service for accessing foundation models from multiple providers (Anthropic, Meta, Mistral, Amazon's own models and more) under AWS contracts.
What the Act means when you deploy AWS (Bedrock)
AWS mainly hosts and distributes third-party models — for most models on Bedrock the GPAI provider obligations sit with the original model provider, while AWS answers for the platform, regions and security. As deployer you must still classify your own use case and meet transparency and oversight duties; the multi-model setup makes it extra important to know which model (and which provider) sits behind each feature.
Questions to send AWS (Bedrock)
Copy these into your supplier questionnaire. Answers in writing beat answers on a call — your Article 26 file needs evidence, not recollection.
- Q1
For each model we use on Bedrock, who is the AI Act GPAI provider, and what model documentation is available to us?
- Q2
Is our prompt and completion data used for model training by AWS or by the model providers, and where is that excluded?
- Q3
Can we keep all Bedrock processing in EU regions, and what does the AWS DPA say about transfers?
- Q4
What logging and retention applies to our Bedrock usage, and who can access it?
Documents to request for your file
- Instructions for use / product documentation for the exact service and version you deploy
- Identification of the underlying model(s) and their provider, plus the documentation passed through to you
- 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. 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 AWS (Bedrock) directly. Public pages change; profiles do not update themselves.
Important: what this page is not
This profile is based on publicly available information about AWS (Bedrock) and describes only who they are and what they offer. It makes no assessment of AWS (Bedrock)’s compliance, contains no rating or certification claim, and is not endorsed by or affiliated with AWS (Bedrock).
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.
Other vendors in Cloud AI platform
Run the free 2-minute checker — deterministic classification, no sign-up.