Amazon Bedrock Guardrails is a set of configurable safeguards for generative AI applications, offered by Amazon Web Services. It filters harmful text and images, blocks topics and words the customer defines, masks or blocks personal data in inputs and responses, checks whether responses are grounded in a source, and uses Automated Reasoning to validate responses against logical rules. Through the ApplyGuardrail API it also works without invoking a model, and it supports self-hosted or third-party models. It is a managed AWS service.
Amazon is a US company, so AWS falls under US jurisdiction, including the CLOUD Act. AWS lets customers choose the Region in which their content is stored, and the Standard Contractual Clauses are part of the AWS Service Terms. With cross-Region inference, requests stay within Regions of the same geography, but input prompts and output results may move outside the primary Region. Which Regions Guardrails supports depends on the model chosen; check this for an EU-only set-up.