> ## Documentation Index
> Fetch the complete documentation index at: https://docs.akhara.ai/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Company name is Akhara AI (never Rubric AI). Keep lowercase rubric/rubrics only when meaning grading criteria.
> Expert Review (docs path talent/) is enterprise BYO experts for audit and review: invite customer specialists; do not pitch Akhara recruiting or a public expert career portal. RLHF and domain writing are secondary work types.
> Prefer concrete API examples against public hosts: Environments eval API https://agi.akhara.ai, Control plane PDP https://api.akhara.dev, Evaluation https://app.akhara.ai / https://api.akhara.ai, Expert Review portal https://talent.akhara.ai.
> Do not invent a public hostname for private orchestrators or env API internals.
> Do not confuse control-plane latches with Environments confirmation latches.
> Environments SDK/API examples: curl against https://agi.akhara.ai. Evaluation SDK: from akhara import Akhara and AKHARA_API_KEY.
> Start with /llms.txt for the docs index and OpenAPI links; fetch individual pages as .md exports.

# Data protection

> AES-256 encryption at rest, TLS 1.2+ in transit, customer-managed keys, redaction of sensitive content, and configurable retention.

Agent traffic is sensitive by construction: prompts, tool arguments, and trajectories routinely carry customer records, cardholder data, or PHI. Akhara treats all of it as regulated data by default.

## Encryption at rest

All stored data is encrypted with AES-256-GCM: primary databases, object storage, search indexes, caches, and backups. Akhara uses envelope encryption, so data encryption keys are themselves encrypted by a master key held in a key management service, and keys rotate without re-encrypting the underlying data.

## Encryption in transit

All network communication uses TLS 1.3, with TLS 1.2 as the enforced minimum and older protocol versions disabled. Cipher suites are limited to AEAD suites with perfect forward secrecy, and HSTS is enabled on all public hosts. Enterprise clients can pin certificates in the SDK.

## Key management

* Managed keys are held in AWS KMS and rotate automatically every year, with on-demand rotation available
* Enterprise workspaces can bring their own keys (BYOK) or use customer-managed keys (CMK), keeping key custody in your own KMS
* In [self-hosted deployments](/security/deployment-models), keys never leave your infrastructure: Akhara components encrypt against the key service you operate

## Sensitive content in logs and traces

Verification requires seeing content, so the platform is built to see it without leaking it:

* PII and PHI detection can redact, tokenize, or hash identifiers before content is stored or routed to reviewers ([PHI/PII handling model](/evaluation/docs/architecture/phi-pii-handling))
* Control plane `WARN` verdicts can rewrite content in-flight, so redacted versions are what actually get delivered ([verdicts](/control-plane/concepts/verdicts))
* Evidence records that retain original and final content are org-partitioned, access-controlled, and encrypted like any other stored data

## Model providers

Decisions the PDP makes about your content are used to produce the verdict and its evidence record, not to train models. Where automated checks call an external LLM provider (for example LLM-judged rubric scoring in Evaluation), enterprise workspaces run those calls under zero-retention terms with the provider. Self-hosted deployments can point the same checks at models you host, so content never reaches a third-party provider at all.

## Retention and deletion

Retention is configurable per data type and workspace. Evidence and audit retention follows your contract, from 30 days on trial up to 7 years for regulated enterprises (see [tenancy and isolation](/security/tenancy)). Deletion requests, including GDPR erasure, are supported with the caveat that audit logs keep their regulatory minimums.