RFI - DEA GenAI Solution (FINAL).pdf

Request for Information (RFI): Internal DEA Generative AI Language Model Platform

Issued By:

U.S. Department of Justice Drug Enforcement Administration (DEA)

RFI Number: D-25-DA-0001 Issue Date: September 4, 2025 Response Due Date: September 25, 2025

1. PURPOSE

The Drug Enforcement Administration (DEA), under the U.S. Department of Justice (DOJ), is seeking information from qualified vendors with experience designing, developing, and deploying secure generative Artificial Intelligence (AI) platforms utilizing Large Language Models (LLMs) for internal federal government use. DEA is looking for both on-premises and/or SaaS solutions. This RFI is issued to inform the potential development and acquisition strategy for a DEA-specific generative AI system capable of operating within a secure environment and aligned with federal cybersecurity, privacy, and AI governance frameworks.

This RFI follows best practices outlined by the Procurement Excellence Network and is structured to promote transparency, stakeholder engagement, and results-driven acquisition planning. It emphasizes the importance of industry insight to inform scoping, requirement setting, and outcome definition. Respondents are encouraged to provide innovative approaches grounded in industry standards, government interoperability, and outcome-based performance.

The DEA aims to leverage generative AI capabilities that enhance operational effectiveness, increase employee productivity, and accelerate knowledge discovery across its workforce. The system development will prioritize support for criminal and administrative (diversion control) investigative responsibilities—core mission functions of the agency—while providing utility to all DEA employees. The envisioned system will enable:

• Automated drafting of first-draft reports, case summaries, and affidavits

• Rapid synthesis of internal and external data to support intelligence and enforcement actions

• Time-saving generation of meeting briefs, summaries, and internal bulletins

• Enhanced collaboration by democratizing access to organizational knowledge

• Augmented decision-making support through evidence-based prompt responses

• Reduced manual burden for repetitive writing and research tasks

• Faster onboarding and training of new personnel via AI-assisted knowledge retrieval

The solution is expected to support Sensitive But Unclassified (SBU), Law Enforcement Sensitive (LES), For Official Use Only (FOUO) and Controlled Unclassified Information (CUI), facilitate compliance with DOJ AI policy directives, and enable scalable secure deployment across the DEA enterprise. It must reflect the complexity, diversity, and global operational footprint of DEA’s investigative and regulatory mission.

2. BACKGROUND

The Drug Enforcement Administration (DEA) is the lead federal agency for enforcing the controlled substances laws and regulations of the United States. As a component of the Department of Justice, DEA's mission includes investigating and dismantling criminal drug networks, regulating the manufacture and distribution of controlled substances, and collaborating with domestic and international partners to combat transnational organized crime.

DEA has more than 15,000 employees worldwide, including special agents, diversion investigators, intelligence research specialists, chemists, and support personnel. These professionals operate across more than 200 domestic offices and nearly 100 foreign offices in over 60 countries. The agency supports criminal and administrative investigations, intelligence gathering and analysis, regulatory enforcement, and policy development to disrupt the illicit drug trade and ensure public health and safety.

As part of its modernization and digital transformation efforts, DEA is evaluating how artificial intelligence, specifically generative AI powered by Large Language Models (LLMs), can support its diverse workforce. The agency recognizes that secure and scalable AI solutions have the potential to enhance investigative documentation, accelerate analytical research, and reduce manual workload across mission areas—including case development, diversion control, legal reviews, strategic intelligence, and training.

Federal agencies are increasingly implementing generative AI systems to streamline operations, reduce administrative burden, and support mission-critical work. Research shows that generative AI tools can significantly improve public sector productivity, enhance content accuracy, and enable faster data interpretation for law enforcement and regulatory activities.

Agencies aligned with national security and law enforcement are leading in adoption due to the high value placed on efficiency, data synthesis, and decision support.

The DEA seeks to explore options for a secure generative AI system capable of:

• Ingesting internal documents and guidance, including policy manuals, investigation SOPs, training materials, and case summaries

• Operating in a DEA-authorized secure environment (e.g., FedRAMP High)

• Functioning within FISMA, NIST, and CJIS standards and frameworks, as well as DOJ/DEA policies and standards, with strict prohibition of foreign access and full compliance with CJIS standards1

• Supporting productivity, content generation, code development, summarization, interoperability, and knowledge discovery for internal mission support from both modern and legacy systems

3. OBJECTIVES

The DEA envisions a Generative AI system that will:

• Be accessible only to DEA employees and contractors with appropriate clearances

• Support the generation of first drafts of investigative and administrative documents— including affidavits, case summaries, and diversion control reports

• Assist with research across internal documents and open-source information to enhance operational analysis and strategic planning

• Generate summaries, meeting briefs, and bulletins for headquarters and field personnel

• Provide recommendations for standard operating procedures based on internal DEA doctrine and precedent

• Offer a searchable and expandable prompt library curated for DEA use cases, tailored to core mission areas such as enforcement, intelligence, policy, training, and compliance

• Ingest and dynamically learn from new DEA documents over time to ensure model outputs remain relevant and aligned with evolving agency practices

• Be procured, governed, and measured through outcomes-based frameworks to ensure that the solution directly supports DEA’s investigative effectiveness, operational readiness, and knowledge retention across its global mission footprint

1 Federal Information Security Modernization Act (FISMA), National Institute of Standards and Technology (NIST), and Criminal Justice Information Services (CJIS)

4. SCOPE OF INFORMATION REQUESTED

In addition to the categories outlined below, respondents are encouraged to highlight how their offerings align with modern government acquisition principles including interoperability, scalability, and agile methodologies. Emphasis should be placed on solutions that reduce time-to-value and support continuous feedback.

A. Technical Capabilities

• LLM platform architecture, hosting, and security controls

• Model fine-tuning or retrieval-augmented generation (RAG) options for DEA-specific knowledge

• Integration with DEA systems, AWS GCC High, or Azure GCC High

• Role-based prompt libraries for mission-specific use

• Auditing, transparency, human review, explainability, and content provenance tracking

• Conducting AI use case testing protocols, including the development of appropriate performance metrics

B. Data Management

• Support data and AI governance technical requirements, processes, and policies (i.e.

access permission controls)

• Secure ingestion of structured/unstructured data

• Access control and dynamic learning mechanisms

• Synthetic test environments and sandboxing capabilities

C. Privacy, Ethics, and Governance

• DOJ and federal compliance, including AI Bill of Rights

• Ethical safeguards for accuracy, bias, and misuse prevention

• Alignment with DOJ guidelines with policy that may be adopted by DEA

D. User Experience and Support

• DEA use-case-based prompt templates and workflows

• Training/onboarding resources

• Feedback loops and user support mechanisms

E. Licensing and Delivery

• Licensing models (user-based, enterprise, usage)

• FISMA, NIST, and CJIS compliance, included but not limited to FedRAMP authorization

• Delivery schedules with clear MVP and full deployment milestones, with an emphasis on rapid prototype development for immediate testing and identification of risks and challenges as appropriate

5. PROMPT LIBRARY REQUIREMENT

The solution must include a DEA-curated prompt library that is:

• Categorized by mission area (e.g., enforcement, intelligence, legal)

• Searchable and expandable with validated use cases

• Governed by user-contribution workflows (e.g., governance board)

• Continuously updated and lifecycle-managed

6. PROCUREMENT STRATEGY AND CONTRACTING RECOMMENDATION

DEA anticipates issuing a Statements of Objectives (SOOs), with DEA’s overarching goals and performance objectives such as:

• Reduction in time spent drafting case reports and affidavits

• Improvement in speed and accuracy of research and brief generation

• Increased adoption and scalability of the government’s institutional knowledge across the enterprise

DEA seeks:

• KPIs tied to productivity and mission impact

• Agile, iterative delivery with user co-design

• Evidence of on-premises and/or SaaS deployment experience with adherence to FISMA, NIST, and CJIS frameworks, standards, and policies

• Interoperability with DOJ and DEA modern and legacy systems

• Alignment with the Federal AI Action Plan, EOs, OMB Memos, and DOJ guidance

• Ethical and responsible AI approach and testing objectives

Proposals should include baselines, milestones, and workload dashboard approaches to monitor and report progress and performance.

7. REQUIRED QUESTIONS FOR RESPONDENTS

Vendors shall answer the following in their response:

1. How will your solution support DEA’s investigative documentation priorities?

2. What strategies ensure secure ingestion and continuous learning from internal content?

3. What specific mechanisms will enforce FISMA, NIST, CJIS, and DOJ/DEA compliance as appropriate?

4. How are foreign interests explicitly restricted technically and procedurally prevented?

5. How can DEA users contribute to or validate prompts?

6. Please describe your data and AI governance framework, including the technical and process solutions to support enterprise compliance, while accelerating AI adoption

7. What are your audit, explainability, and provenance tracking capabilities?

8. How will your approach demonstrate measurable mission-aligned outcomes?

9. Describe system interoperability between secure environments and modern/legacy systems.

10. What is your user onboarding and training support model?

11. Outline a phased delivery schedule with MVP and full capability milestones.

12. How would you structure outcomes-based metrics to align with investigative and administrative productivity improvements?

13. What performance indicators or logic models do you recommend for evaluating system success?

14. What experience do you have executing outcomes-based contracts with other federal agencies, and what lessons learned would you apply?

15. How does your solution support automating the identification and assignment of meta-data (e.g. cataloguing, ontology, geocoding), aligned to USG, IC, and DOJ data governance and management standards?

16. What is the plan for your solution to attain the standard DOJ and DEA must adhere to, in the event of utilizing output in the support of an investigation or prosecution?

8. SUBMISSION INSTRUCTIONS

Please email the following documents in PDF format in 11-point font, single-spaced to Robert A.

Jones. [Robert.A.JonesJr@dea.gov] by September 25, 2025:

• Cover page, including company name, point of contact, UEI, and contract vehicle info (e.g., GSA MAS), socioeconomic status, and rough order of magnitude (1 page max) o Socioeconomic status (e.g., 8(a), HUBZone, SDVOSB, WOSB) o Rough Order of Magnitude (ROM)

• Capability Statement (10 pages max)

• Responses to Section 4 and Section 7 questions (15 pages max)

• Relevant past performance summaries and Contract Performance Assessment Reports (CPARs) (4 pages max)

Please do not include appendices. Any included visuals are counted towards the page limit.

9. DISCLAIMER

This is a Request for Information (RFI) only. It does not commit the government to issue a solicitation, award a contract, or reimburse any costs associated with responses. Information submitted will be used solely for planning purposes and may inform future procurement strategies.