US Department of Homeland Security

Classification & Impact

CAIO Determination / Justification 2025

The Chief AI Officer (CAIO) reviewed each use case and determines whether it qualifies as "high-impact" under OMB guidance. The justification field explains the reasoning behind that determination. This replaced the separate rights-impacting and safety-impacting classifications used in the 2024 Inventory July Revision.

High-Impact AI 2025

An AI use case that could significantly affect the rights, safety, or civil liberties of individuals. Under the 2025 Inventory schema, use cases are classified as "High-impact," "Presumed high-impact but determined not high-impact," or "Not high-impact."

Presumed High-Impact 2025

A use case that falls into a category presumed to be high-impact (such as law enforcement or immigration screening), but which the CAIO has reviewed and determined does not meet the threshold for the high-impact designation, with justification provided.

Rights-Impacting 2024 Rev

A classification used in the 2024 Inventory July Revision indicating whether an AI use case could affect individual rights, such as due process, privacy, or equal protection. Replaced by the "high-impact" framework in the 2025 Inventory.

Safety-Impacting 2024 Rev

A classification used in the 2024 Inventory July Revision indicating whether an AI use case could affect physical safety, critical infrastructure, or emergency response. Replaced by the "high-impact" framework in the 2025 Inventory.

AI Classification 2025

Categorizes the type of AI technology used: Agentic AI (systems that take autonomous actions), Generative AI (systems that produce new content), or other AI/ML approaches.

Agentic AI

AI systems capable of taking autonomous actions in the real world with limited human oversight, such as making decisions, executing tasks, or interacting with external systems on behalf of a user or organization.

Lifecycle & Status

Stage (2025 Inventory)

Pre-deployment — Under development or testing, not yet in operational use.
Pilot — Being tested in a limited operational setting.
Deployed — Fully operational and in active use.
Retired — No longer in use; decommissioned.

Stage (2024 Inventory July Revision)

Initiated — Concept or planning phase.
Acquisition/Development — Being procured or built.
Implementation/Assessment — Being deployed and evaluated.
Operation & Maintenance — Fully operational with ongoing support.
Retired — No longer in use.

Security & Compliance

ATO (Authority to Operate)

A formal authorization granted by a federal agency's authorizing official that permits an IT system to operate. Indicates the system has undergone a security assessment and the associated risks have been accepted.

PIA (Privacy Impact Assessment)

A formal analysis of how personally identifiable information (PII) is collected, stored, shared, and protected within a system. Required by the E-Government Act when a federal system handles PII.

PII (Personally Identifiable Information)

Information that can be used to identify an individual, such as name, Social Security number, biometric data, or other distinguishing characteristics.

SAOP (Senior Agency Official for Privacy) 2024 Rev

The senior official designated by DHS to oversee privacy compliance and policy. The 2024 Inventory July Revision tracks whether the SAOP has assessed each AI use case for privacy implications.

Procurement & Infrastructure

Procurement Method

How the AI capability was acquired. Common values include commercially procured (purchased from a vendor), developed in-house by government staff, developed under contract, or open-source.

Vendor

The commercial entity providing the AI technology or service. This field is populated when the AI capability was procured from an external provider rather than built in-house.

HISP (High Impact Service Provider) 2024 Rev

A designation under FISMA (Federal Information Security Modernization Act) for systems or services that are critical to the functioning of the federal government. The 2024 Inventory July Revision tracks whether each AI use case supports a HISP-designated service.

Oversight & Risk Management

Impact Assessment

An evaluation of the potential effects of an AI system on individuals, communities, or operations. May include algorithmic impact assessments, civil rights assessments, or other formal reviews.

Independent Review

Whether the AI system has undergone evaluation by a reviewer not involved in its development, such as an inspector general, external auditor, or independent testing body.

Ongoing Monitoring

Continuous or periodic evaluation of an AI system's performance, accuracy, bias, and compliance after deployment.

Appeal Process

Whether individuals affected by an AI-assisted decision have a mechanism to challenge or seek review of that decision.

Fail-Safe 2025

Mechanisms built into an AI system to ensure safe behavior in the event of a malfunction, unexpected input, or system failure.

Other Terms

Bureau

The DHS sub-agency or component operating the AI use case, such as CBP (Customs and Border Protection), ICE (Immigration and Customs Enforcement), TSA (Transportation Security Administration), CISA (Cybersecurity and Infrastructure Security Agency), or USCIS (U.S. Citizenship and Immigration Services).

Topic Area

The functional category of the AI use case, such as law enforcement, cybersecurity, immigration, disaster response, or administrative operations.

Demographic Variables

Whether the AI system uses or processes data related to demographic characteristics such as race, ethnicity, gender, or age. Tracking this field helps assess potential bias risks.