From the AWS edition of October 30, 2026
AWS publishes a CJIS-oriented RAG reference architecture on Bedrock for municipal law enforcement intelligence units, priced under $200 a month at POC scale
Amazon Web Services (AWS Public Sector) · Public safety / law enforcement · United States (state and local)
- Publisher
- AWS Public Sector Blog
- Original publication
- 2026-10-27
- Source retrieved
- 2026-10-31
What happened
An AWS solutions architect who works with state and local agencies describes a retrieval-augmented generation (RAG) tool on Amazon Bedrock Managed Knowledge Base. It lets law enforcement analysts ask natural-language questions of case files and CJIS records and get cited answers. The post argues that CJIS Security Policy v6.1 does not mandate a specific provider or FedRAMP level, so commercial us-west-2 is viable when CJIS is the main driver. GovCloud is recommended when FedRAMP High, ITAR or IL4/IL5 also apply. Controls include customer-managed KMS encryption, CloudTrail, full prompt-response invocation logging, and ACL-aware retrieval that fails closed. AWS estimates $78-$191 per month for about 5 GB of documents and 2,000 queries.
Why it matters
Directly relevant to municipal and county police intelligence units and state fusion centers that handle CJI. It also matters to state CJIS Systems Agencies, which the post says assess compliance. Cross-tagged to public-safety because the use case is law-enforcement records.
Evidence and measured results
Vendor-authored reference design. The 'municipal intelligence unit' is a hypothetical scenario, and no agency deployment, retrieval-accuracy test, or user study is reported. The cost table is an AWS estimate for a proof-of-concept workload (about 5 GB, 2,000 queries a month) using list prices, including $4.00 per 1,000 agentic retrieval calls versus $1.00 for standard Retrieve. Regulatory statements (CJIS v6.1, June 2026; the 2021 shift to a technical-controls model) are AWS's characterization.
Limitations and uncertainty
Vendor guidance with no measured outcomes. Cost figures are estimates at small scale and exclude application build, identity integration and staffing. Commercial-region viability depends on each state CSA's interpretation and the CJIS Security Addendum. The excerpt is AWS's own disclaimer, not an independent ruling.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-30; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Analysts manually search large volumes of case files and CJIS records and are unsure whether generative AI can be applied to law enforcement data.
- Stakeholders
- Police or sheriff intelligence unit commander
- Agency CIO/IT security lead
- State CJIS Systems Agency (CSA) liaison
- Records management supervisor
- City/county attorney
- Discovery questions
- Has your state CSA taken a position on CJI in commercial AWS regions versus GovCloud?
- What document volumes and query volumes would a pilot cover?
- How are analyst identities managed today, and can they be passed reliably to an application?
- What log retention period does your CJIS policy require?
- Value hypothesis
- For a small corpus, a cited natural-language retrieval tool could shorten analyst search time at a modest monthly cloud cost, if access controls and the CSA review hold up.
- Potential engagement
- A 6-8 week pilot on a sanitized or non-CJI corpus to measure retrieval precision, citation accuracy and access-control behavior before seeking CSA review.
- Unsupported claims
- That the architecture is CJIS-compliant out of the box
- Any measured analyst time savings
- Production cost at agency scale
- Accuracy of answers on real case files
Pre-sales engineering
Role takeaway
- Fit
- Good fit for agencies already on AWS with a modest document corpus and clear access rules. Weak fit where identity data is fragmented.
- Architecture and integration
- S3 ingestion bucket with customer-managed KMS, a Bedrock Managed Knowledge Base with ACL-aware agentic retrieval, a foundation model, and CloudTrail plus invocation logging. The UI comes later through an AgentCore Gateway agent.
- Prerequisites
- Authoritative identity source with stable user identifiers
- Document-level permission metadata in the source system
- CSA engagement and CJIS Security Addendum
- Log retention and access policy for prompt-response logs
- Constraints
- CJIS interpretation varies by state CSA
- ACL cache delay of up to one hour on revocations
- GovCloud needed if FedRAMP High/ITAR/IL4-5 also apply
- Feature availability may differ between commercial and GovCloud regions
- Security
- The knowledge base filters but does not authorize; the application must authenticate users. Encrypt with customer-managed keys, restrict and monitor the invocation log store, and test fail-closed behavior.
- Proof of value
- Measure retrieval precision and citation accuracy on a labeled question set
- Test that a revoked user loses access, and record the time it takes
- Confirm ACL errors return zero results
- Compare actual monthly cost against the $78-$191 estimate at pilot volume
Delivery
Role takeaway
- Work
- Stand up the pilot environment, ingest a scoped corpus, integrate identity, configure logging and retention, and run analyst evaluation sessions.
- Dependencies
- CSA review and approval path
- Identity provider integration
- Records unit time to label test questions
- AWS account with required region and KMS setup
- Ownership
- Intelligence or records unit owns content and access rules; agency IT owns the AWS environment and identity; the CSA liaison owns compliance evidence.
- Skills and adoption
- Train analysts to verify cited sources and report bad answers. Train IT on Bedrock Knowledge Base ACL configuration and log handling.
- Governance checkpoints
- CSA consultation before indexing live CJI
- Access-control test sign-off
- Log retention policy approval
- Go/no-go review on measured retrieval quality
- Acceptance criteria
- Proposed: at least 90% of answers on the test set cite the correct source document
- Proposed: revoked access takes effect within the documented cache window in 100% of tests
- Proposed: zero unauthorized documents returned in access-control test suite
- Proposed: monthly pilot cost within agreed budget
- Risks
- Over-reliance on generated summaries in investigations
- Access revocation lag exposing compartmented cases
- Sensitive data accumulating in invocation logs
- CSA rejecting commercial-region hosting after build
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Serverless S3 → Bedrock Managed Knowledge Base → foundation model (Claude Sonnet 5 in the estimate). The same template deploys to commercial or GovCloud. Agentic retrieval costs four times standard retrieval, so query volume drives cost. The post recommends a phased path: console testing, then MCP/CLI testing, then production through an AgentCore Gateway agent.
Governance
Who approves, reviews and stays accountable for outcomes?
The post states that no central body certifies CJIS compliance; the state CSA assesses it, and agencies still need policies, personnel controls and AWS agreements. Region choice (commercial vs GovCloud) should be decided by the agency's full compliance picture, not by price alone.
Security and privacy
What data, permissions and controls need testing?
ACL-aware retrieval filters results but is explicitly not authorization: the application must authenticate users and keep identity mappings current. ACL updates can be cached for up to an hour, so revocations lag. Full prompt-response logging creates a sensitive log store holding CJI that needs its own retention policy and access control (the post suggests 5-7 years).
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Analysts need training to treat citations as leads to verify rather than conclusions. Small agencies may lack staff to run identity and ACL lifecycles, which the design depends on.
Procurement
What should contracts, pricing and exit terms secure?
The low POC price point may let agencies pilot below formal procurement thresholds. Buyers should still require a CJIS Security Addendum, CSA review and a defined exit path for indexed data and logs.
Operating model
Which teams own the service once it runs?
Ownership is split between the records/intelligence unit (content and access rules), IT (identity and AWS), and the CSA liaison (compliance evidence). Revocation procedures must account for the cache delay.
Publication history
- 2026-10-30AWS · Issue 027 resources
Stable resource ID: aws-bedrock-cjis-rag-reference-architecture-20261027