leeroy-jankins/DoD-Instruction-8170-01-Online-Information-Management-And-Electronic-Messaging

📚 DoD Instruction 8170.01 Online Information Management and Electronic Messaging

• Maintainer: Terry Eppler
• Ownership: U.S. Department of Defense

---

📋 Overview

Dataset Summary

The DoD Instruction 8170.01 Online Information Management and Electronic Messaging Dataset is a structured natural-language question-answering dataset derived from DoD Instruction 8170.01, Online Information Management and Electronic Messaging.

DoD Instruction 8170.01 establishes Department of Defense policy, assigns responsibilities, and prescribes procedures for conducting, establishing, operating, and maintaining electronic messaging services used to collect, distribute, store, present, and otherwise process official DoD information. It also governs the management of official DoD information on the DoD Information Network and other online environments.

The source instruction addresses online information management, electronic mail, websites, social media, external official presences, public web services, privacy, cybersecurity, records management, accessibility, information quality, web analytics, mobile optimization, web application programming interfaces, hyperlinks, digital signatures, encryption, official and personal account use, and web measurement and customization technologies.

The dataset contains question-answer records designed to support language-model training, retrieval-augmented generation, policy question answering, compliance analysis, information extraction, summarization, and domain adaptation involving DoD online information and electronic messaging requirements.

Each record contains a natural-language policy question and a detailed answer grounded in the source instruction. Answers are written as complete explanations rather than simple citations or isolated quotations. The questions avoid directly referring to the source document and are intended to resemble realistic questions that may be asked by DoD personnel, system owners, public-affairs professionals, privacy officials, records managers, cybersecurity personnel, website administrators, and policy analysts.

Supported Tasks

Retrieval-Augmented Generation

The dataset may be used to create vector, lexical, or hybrid retrieval systems that identify relevant policy guidance concerning:

• official DoD information distributed online;
• DoD electronic messaging and electronic messaging services;
• public and private DoD websites;
• official and personal electronic messaging accounts;
• external official presences;
• non-DoD-controlled electronic messaging services;
• website accessibility;
• advertising and endorsement;
• cybersecurity and transport-layer security;
• digital signatures and encryption;
• web application programming interfaces;
• privacy notices and Privacy Act statements;
• privacy impact assessments;
• personally identifiable information;
• records management;
• information-quality complaints;
• website registration;
• search functionality;
• mobile optimization;
• multilingual content;
• web measurement and customization technologies;
• social-media account archiving; and
• official online communication responsibilities.

Question Answering

The dataset supports closed-domain question answering in which answers are grounded in the requirements, responsibilities, procedures, definitions, and administrative mechanisms established by DoD Instruction 8170.01.

Example questions include:

• Which organizations and activities are subject to the online-information requirements?
• Which electronic-messaging activities are excluded from the instruction?
• Why should electronic messaging services be designed around data rather than documents?
• When must controlled unclassified electronic messages be encrypted?
• What alternative must personnel use when ordinary encryption is unavailable?
• May official messages be automatically forwarded to a personal account?
• What responsibilities are assigned to the DoD Chief Information Officer?
• What role does the Defense Information Systems Agency perform?
• What must Component heads do when an electronic messaging service remains noncompliant?
• How frequently must electronic messaging services be assessed?
• What accessibility requirements apply to official online information?
• What advertising and endorsement restrictions apply to public DoD services?
• What information must appear on a DoD website contact page?
• Which Internet domain should be used as the primary address for DoD-controlled services?
• What disclaimer must accompany hyperlinks to non-U.S. Government websites?
• Which hyperlinks and content are mandatory on principal public DoD websites?
• How may non-DoD-controlled services be used for official communication?
• Under what conditions may a personal account be used for an official message?
• When must a Privacy Act statement be displayed?
• When is a privacy advisory required?
• When must a privacy impact assessment be completed?
• How quickly must a record sent through a nonofficial account be forwarded to an official account?
• What restrictions apply to web measurement and customization technologies?
• What distinguishes Tier 1, Tier 2, and Tier 3 technologies?
• How may an affected person request correction of publicly distributed information?
• What appeal rights apply to an information-quality determination?
• What information must be reported annually concerning information-quality complaints?

Summarization

The dataset can support summarization of:

• general policy requirements;
• organizational responsibilities;
• cybersecurity procedures;
• privacy procedures;
• records-management requirements;
• public-website standards;
• external official presence requirements;
• official and personal account restrictions;
• information-quality principles;
• complaint-resolution procedures;
• web-measurement requirements;
• glossary terms; and
• referenced statutory and regulatory authorities.

Text Classification

The dataset may support classification by policy area, including:

• applicability;
• information-centric architecture;
• electronic messaging;
• public websites;
• social media;
• cybersecurity;
• operations security;
• accessibility;
• privacy;
• personally identifiable information;
• controlled unclassified information;
• records management;
• public affairs;
• information quality;
• mobile optimization;
• web analytics;
• digital signatures;
• encryption;
• hyperlinks;
• application programming interfaces;
• data management;
• website registration;
• search and discoverability;
• advertising;
• copyright;
• official account use;
• personal account use; and
• web measurement and customization technologies.

Named Entity Recognition and Information Extraction

The dataset may be used to extract entities and relationships such as:

• DoD organizations and senior officials;
• assigned responsibilities;
• policy requirements;
• prohibited activities;
• required approvals;
• compliance deadlines;
• reporting deadlines;
• information-security categories;
• privacy requirements;
• records-management requirements;
• statutory and regulatory authorities;
• executive orders;
• Office of Management and Budget memoranda;
• DoD directives, instructions, and manuals;
• required website notices;
• required hyperlinks;
• official online services;
• external official presences;
• web-technology tiers;
• information-quality attributes;
• complaint procedures;
• appeal procedures; and
• defined policy terms.

Dataset Structure

The dataset contains one policy question-answer pair per record.

Each question is phrased as an authentic user inquiry rather than as a request to identify a page, paragraph, or location in the source instruction. Each answer provides a substantive explanation grounded in the instruction and generally contains at least three complete sentences.

Dataset Fields

Field	Type	Description
ID	integer	Sequential unique identifier for the dataset record.
User	string	Role label identifying the question as a user message.
Question	string	Natural-language question grounded in DoD Instruction 8170.01.
Agent	string	Role label identifying the response as an agent message.
Answer	string	Detailed answer based on the source instruction.

Example Record

{
  "ID": 1,
  "User": "User",
  "Question": "Which organizations and activities are subject to the online-information and electronic-messaging requirements?",
  "Agent": "Agent",
  "Answer": "The requirements apply to OSD, the Military Departments, the Joint Staff, the Combatant Commands, the Office of Inspector General of the Department of Defense, the Defense Agencies, the DoD Field Activities, and other organizational entities within the Department. They also apply to official DoD information online, DoD electronic messaging, and DoD electronic messaging services, including qualifying services operated or used by non-DoD entities. Contractors and other non-DoD organizations are covered to the extent established by the contract or other instrument authorizing their mission support or access."
}

Conversation Format

Each row can also be converted into a conversational structure for supervised fine-tuning:

{
  "messages": [
    {
      "role": "user",
      "content": "Which organizations and activities are subject to the online-information and electronic-messaging requirements?"
    },
    {
      "role": "assistant",
      "content": "The requirements apply to OSD, the Military Departments, the Joint Staff, the Combatant Commands, the Office of Inspector General of the Department of Defense, the Defense Agencies, the DoD Field Activities, and other organizational entities within the Department. They also apply to official DoD information online, DoD electronic messaging, and DoD electronic messaging services, including qualifying services operated or used by non-DoD entities. Contractors and other non-DoD organizations are covered to the extent established by the contract or other instrument authorizing their mission support or access."
    }
  ]
}

Dataset Creation

Source Data

The dataset is derived from:

DoD Instruction 8170.01, Online Information Management and Electronic Messaging

• Originating Component: Office of the Chief Information Officer of the Department of Defense
• Original Effective Date: January 2, 2019
• Change 2 Effective Date: March 12, 2025
• Releasability: Cleared for public release
• Source Organization: U.S. Department of Defense
• Document Type: Department of Defense Instruction

The instruction incorporates and cancels:

• DoD Instruction 8550.01, DoD Internet Services and Internet-Based Capabilities, September 11, 2012; and
• Deputy Secretary of Defense Memorandum, Ensuring Quality of Information Disseminated to the Public by the Department of Defense, February 10, 2003.

Source Coverage

The dataset covers the principal substantive portions of the instruction, including:

• Section 1, General Issuance Information;
• Section 2, Responsibilities;
• Section 3, Procedures;
• Appendix 3A, Ensuring the Quality of Information Distributed to the Public; and
• relevant glossary definitions.

The procedural coverage includes:

• accessibility;
• advertising and endorsement;
• annual assessments;
• archiving official social-media accounts;
• branding;
• cloud services;
• online information collection;
• copyright;
• cybersecurity;
• transport-layer security;
• datasets and web APIs;
• Digital Analytics Program implementation;
• digital signatures;
• website contact information;
• Internet domains;
• encryption;
• federal information systems;
• image alteration;
• information control, distribution, and marking;
• hyperlinks;
• mobile code;
• mobile optimization;
• multilingual content;
• official use of non-DoD-controlled services;
• plain writing;
• personal use of non-DoD-controlled services;
• Privacy Act statements;
• privacy advisories;
• privacy impact assessments;
• privacy incidents;
• public-website standards;
• records management;
• registration;
• search and discoverability; and
• web measurement and customization technologies.

Processing Pipeline

The dataset-generation process included:

1. Extracting text from the publicly releasable source instruction.
2. Reviewing the instruction’s sections, subsections, figures, appendix, glossary, and procedural requirements.
3. Removing repeated page headers, page numbers, table-of-contents artifacts, and extraction noise.
4. Identifying substantive policy rules, responsibilities, conditions, prohibitions, exceptions, deadlines, and administrative procedures.
5. Generating natural-language questions that can be answered from the source material.
6. Avoiding questions that merely ask where information appears in the instruction.
7. Producing explanatory answers containing sufficient policy context.
8. Preserving distinctions among mandatory requirements, recommendations, conditions, exceptions, and discretionary authorities.
9. Verifying that each answer is supported by the source instruction.
10. Formatting the final records using the ID, User, Question, Agent, and Answer schema.

Question-Generation Strategy

Questions were designed to test understanding of:

• policy purpose and scope;
• applicability and exclusions;
• organizational responsibilities;
• mandatory and discretionary actions;
• prohibited conduct;
• approval and registration requirements;
• cybersecurity controls;
• privacy controls;
• records-management duties;
• public-release requirements;
• website and digital-service standards;
• use of non-DoD services;
• employee use of personal accounts;
• complaint-resolution procedures;
• administrative deadlines;
• information-quality standards; and
• relationships among multiple requirements.

The questions were written to support reasoning and policy interpretation rather than simple passage matching.

Answer-Generation Strategy

Answers were written to:

• directly answer the question;
• explain the governing rule;
• identify relevant conditions or exceptions;
• describe implementation consequences;
• distinguish related requirements when necessary;
• use terminology consistent with the instruction; and
• provide enough context to be useful without access to the original source passage.

Answers do not intentionally introduce legal authorities, requirements, or procedures that are absent from the source instruction.

Data Splits

The dataset is distributed as a single set of 150 records.

Recommended Split Design

Split	Purpose	Records
train	Model training, retrieval indexing, or domain adaptation.	120
validation	Prompt development, retrieval evaluation, or model selection.	15
test	Held-out evaluation.	15

A deterministic 80/10/10 split may be created using the ID field.

splits:
  - name: train
    num_examples: 120
  - name: validation
    num_examples: 15
  - name: test
    num_examples: 15

For retrieval-augmented generation, all 150 records may instead be used as a single corpus split.

splits:
  - name: corpus
    num_examples: 150

Loading the Dataset

Using the Hugging Face Datasets Library

from datasets import load_dataset

dataset = load_dataset(
    "csv",
    data_files={
        "train": "dataset.csv"
    }
)

print(dataset["train"][0])

Loading the Markdown Table with Pandas

The preferred machine-readable format is CSV. If the dataset is distributed as a Markdown table, it should first be converted to CSV or parsed using an appropriate table-processing library.

Loading the CSV with Pandas

import pandas as pd

dataset = pd.read_csv("dataset.csv")

print(dataset.head())
print(dataset.columns.tolist())

Expected columns:

[
    "ID",
    "User",
    "Question",
    "Agent",
    "Answer"
]

Converting to Chat Messages

import pandas as pd

dataset = pd.read_csv("dataset.csv")

conversations = [
    {
        "messages": [
            {
                "role": row["User"].lower(),
                "content": row["Question"]
            },
            {
                "role": "assistant",
                "content": row["Answer"]
            }
        ]
    }
    for _, row in dataset.iterrows()
]

print(conversations[0])

When converting the Agent field to a standard chat role, use assistant rather than agent if required by the target model or training framework.

Intended Uses

Primary Intended Uses

This dataset is intended for:

• training and evaluating language models on DoD online-information policy;
• building retrieval-augmented generation systems for DoD electronic messaging guidance;
• developing question-answering systems for public websites, social media, privacy, cybersecurity, records management, and information quality;
• testing policy-grounded response generation;
• creating embeddings for DoD information-management requirements;
• evaluating whether models distinguish requirements, recommendations, exceptions, and prohibitions;
• supporting regulatory and policy text classification;
• extracting organizational roles and responsibilities;
• identifying compliance conditions and deadlines;
• supporting domain adaptation for government information-management applications; and
• researching natural-language processing over public DoD policy documents.

Example Use Cases

• A website administrator asks which notices and hyperlinks must appear on a public DoD website.
• A public-affairs office asks how to establish an external official presence.
• A records manager asks how messages created through a personal account must be preserved.
• A privacy official asks whether a Privacy Act statement, privacy advisory, or privacy impact assessment is required.
• A cybersecurity official asks which certificate validation requirements apply to an external-facing service.
• A Component CIO evaluates a noncompliant electronic messaging service.
• A system owner asks whether a service must be registered or assessed annually.
• A developer asks whether official data should be exposed through a web API.
• A content manager asks how external hyperlinks must be reviewed and disclaimed.
• A policy analyst examines the information-quality complaint and appeal process.
• A data scientist creates embeddings for policy retrieval.
• A model developer evaluates grounded answers concerning official and personal account use.

Out-of-Scope Uses

The dataset should not be used as the sole authority for:

• legal advice;
• final cybersecurity authorization decisions;
• classification determinations;
• controlled-unclassified-information determinations;
• records-disposition decisions;
• Privacy Act compliance determinations;
• privacy-impact-assessment approvals;
• public-release authorization;
• operations-security review;
• approval of an external official presence;
• approval of a non-DoD-controlled electronic messaging service;
• incident-response decisions;
• disciplinary actions;
• contract interpretation;
• official DoD policy determinations; or
• operational decisions involving classified or nonpublic information.

Users should consult the current authoritative version of DoD Instruction 8170.01, controlling statutes, applicable DoD issuances, Component-level policy, legal counsel, privacy officials, records managers, cybersecurity officials, public-affairs officials, and other responsible authorities before making compliance-sensitive decisions.

Licensing and Use Restrictions

DoD Instruction 8170.01 is a publicly released U.S. Government policy document.

The source document is identified as cleared for public release and is available through the DoD Directives Division. U.S. Government works are generally not protected by copyright within the United States when prepared by federal employees as part of their official duties. However, the source may include referenced material, trademarks, graphics, quotations, or other content subject to separate rights or restrictions.

Recommended dataset license field:

license: other

Suggested license statement:

This dataset is derived from a publicly released U.S. Department of Defense instruction. Users should verify the status of source material and any restrictions applicable to generated annotations, derived question-answer pairs, third-party content, graphics, trademarks, or redistributed source text.

The generated questions, answers, formatting, and annotations may be subject to the dataset maintainer’s chosen license, provided that the license does not mischaracterize the legal status of the underlying U.S. Government source.

Data Quality

Strengths

• Contains 150 structured, policy-grounded question-answer records.
• Covers organizational responsibilities and operational procedures.
• Includes public-website, privacy, cybersecurity, records, and social-media requirements.
• Contains detailed answers rather than isolated quotations.
• Preserves distinctions among mandatory, recommended, prohibited, and discretionary actions.
• Includes conditions, exceptions, deadlines, and approval requirements.
• Uses a simple schema suitable for CSV, Pandas, Hugging Face Datasets, and supervised fine-tuning.
• Supports retrieval, question answering, classification, summarization, and information extraction.
• Uses realistic questions rather than questions about page or paragraph locations.
• Is based on a publicly releasable Department of Defense policy source.

Validation Considerations

Dataset maintainers should verify:

• that all 150 identifiers are unique;
• that ID is consistently represented as an integer;
• that every User value is User;
• that every Agent value is Agent;
• that no row contains an empty question or answer;
• that Markdown and CSV versions contain the same records;
• that embedded commas and quotation marks are correctly escaped in CSV;
• that answer text does not contain extraction artifacts;
• that questions do not inadvertently disclose the answer;
• that answers remain grounded in the source instruction; and
• that the source version and effective date are accurately recorded.

Known Limitations

• The dataset represents a single DoD instruction rather than the entire body of DoD information-management policy.
• Many requirements incorporate other statutes, executive orders, OMB memoranda, DoD directives, instructions, and manuals by reference.
• The dataset does not reproduce the complete text of every incorporated authority.
• Generated answers summarize and explain policy but are not substitutes for the controlling source.
• Some requirements depend on Component-specific implementation, approval procedures, contractual language, mission needs, system architecture, or information sensitivity.
• The instruction may be amended, replaced, administratively updated, or supplemented after the source version used for dataset creation.
• The 150 records are designed for coverage and instructional value but do not represent every possible question that could be generated from the source.
• Some source provisions contain complex cross-references that may require review of additional authorities for complete legal or operational interpretation.
• The dataset does not contain classified, controlled unclassified, operational, or Component-internal implementation information.

Bias, Risks, and Limitations

The dataset reflects the policy, terminology, organizational structure, and legal framework of the U.S. Department of Defense.

Potential risks include:

• treating generated answers as official DoD interpretations;
• relying on a summarized answer without reviewing the controlling provision;
• using an outdated version after a policy amendment;
• overlooking incorporated statutes, regulations, memoranda, or Component guidance;
• confusing public and private electronic messaging services;
• confusing official and personal use;
• failing to recognize information sensitivity or classification by compilation;
• assuming that a publicly accessible commercial service is approved for official use;
• overlooking privacy, records, public-release, or operations-security requirements;
• treating recommended practices as mandatory requirements or mandatory requirements as discretionary;
• failing to distinguish a Privacy Act statement from a privacy advisory;
• failing to distinguish Tier 1, Tier 2, and Tier 3 web-measurement technologies; and
• using model output to make compliance decisions without qualified review.

Models trained or evaluated with this dataset should:

• identify uncertainty;
• preserve policy terminology;
• distinguish mandatory and advisory language;
• recognize exceptions and conditions;
• avoid inventing approvals or authorities;
• cite controlling source passages when used in operational applications; and
• refer compliance-sensitive matters to responsible DoD officials.

Personally Identifiable Information

The source instruction is publicly released and is not expected to contain private individual-level records.

The dataset does not intentionally contain:

• personal addresses;
• personal telephone numbers;
• personal e-mail addresses;
• Social Security numbers;
• medical records;
• financial-account information;
• personnel records; or
• other private individual-level data.

The instruction discusses personally identifiable information, Privacy Act systems of records, privacy incidents, privacy impact assessments, and related safeguards as policy subjects. References to PII in the dataset are therefore explanatory and do not represent the inclusion of actual personal data.

Dataset maintainers should nevertheless inspect generated and transformed files before publication to ensure that no unintended metadata or personal information was introduced.

Security Considerations

The dataset is derived from public policy information and should contain no classified or controlled unclassified information.

The dataset should not be combined with:

• classified information;
• controlled unclassified information;
• nonpublic operational information;
• operations-security indicators;
• personally identifiable information;
• protected health information;
• law-enforcement-sensitive information;
• intelligence information;
• authentication credentials;
• system vulnerabilities;
• private communications; or
• Component-internal security documentation,

unless the resulting system and dataset are operated under the appropriate security, privacy, access-control, and records-management requirements.

Public AI or machine-learning services should not be used to process protected DoD information merely because this public dataset concerns DoD policy.

Maintenance

Dataset Maintainer

• Maintainer: Terry Eppler
• Source Organization: U.S. Department of Defense
• Source Issuance: DoD Instruction 8170.01

Update Frequency

The dataset should be reviewed whenever:

• DoD Instruction 8170.01 is changed, reissued, canceled, or superseded;
• a new administrative change becomes effective;
• incorporated DoD issuances are materially revised;
• relevant OMB website, privacy, or digital-service guidance changes;
• public-website standards change;
• cybersecurity or public-key-enabling requirements change;
• privacy or records-management requirements change;
• official social-media guidance changes; or
• errors are identified in the dataset.

Versioning Recommendation

Use semantic versioning tied to source updates and dataset-processing changes.

v1.0.0 - Initial release containing 150 question-answer records.
v1.1.0 - Corrected wording, metadata, or formatting without changing the source edition.
v1.2.0 - Added policy-topic labels, source-section metadata, or retrieval fields.
v1.3.0 - Added validated citations or paragraph references.
v1.4.0 - Added conversational and instruction-tuning configurations.
v2.0.0 - Regenerated the dataset following a substantive revision of DoD Instruction 8170.01.

Each release should record:

• dataset version;
• source-document title;
• source effective date;
• source change number;
• source change effective date;
• record count;
• schema;
• processing changes;
• corrected records; and
• validation status.

Citation

When using this dataset, cite both the dataset release and the source instruction.

Dataset Citation

@misc{dodi_8170_01_online_information_management_dataset,
  title        = {DoD Instruction 8170.01 Online Information Management and Electronic Messaging Dataset},
  author       = {Eppler, Terry},
  year         = {2026},
  howpublished = {Hugging Face Dataset},
  note         = {A 150-record question-answer dataset derived from DoD Instruction 8170.01}
}

Source Citation

@misc{department_of_defense_2019_dodi_8170_01,
  title        = {DoD Instruction 8170.01: Online Information Management and Electronic Messaging},
  author       = {{U.S. Department of Defense}},
  year         = {2019},
  month        = {January},
  note         = {Effective January 2, 2019; Change 2 effective March 12, 2025}
}

Dataset Card Authors

The dataset card was prepared for a document-derived Hugging Face dataset based on DoD Instruction 8170.01, Online Information Management and Electronic Messaging.

Acknowledgements

This dataset is based on a publicly released issuance of the U.S. Department of Defense.

The source instruction was issued by the Office of the Chief Information Officer of the Department of Defense and establishes Department-wide policy and procedures for online information management and electronic messaging.