AI-Powered Search and Recovery

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not high impact: This AI system does not meet the criteria of any of the six pillars that make up High Impact AI in Memorandum M-25-21.

Start date

2026-01-01 (day precision)Source: 01/01/2026 00:00:00

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Cybersecurity

Operational functions

Chatbot or virtual assistant, Summarization, Cyber threat detection Derived by keyword rules; see methodology

Problem the AI is intended to solve

During cloud migration (AWS EVS) and disaster recovery operations, locating specific data across backup snapshots, identifying clean recovery points, and performing forensic analysis of backup data requires significant manual effort and specialized knowledge of backup infrastructure.

Expected benefits

AI-powered conversational search across enterprise backup data enables faster data discovery during recovery operations, reduces time-to-recovery during cyber incidents, and supports forensic investigation by allowing natural language queries against unstructured backup data. SaaS deployment model reduces operational overhead for backup infrastructure management during AWS EVS migration.

System outputs

AI assistant providing natural language search, classification, and summarization across Cohesity Data Cloud backup data. Includes: (1) Conversational search — natural language queries against unstructured data in backup snapshots, (2) AI-powered threat detection — continuous anomaly scanning of backup data for ransomware indicators and insider threats, (3) Recovery intelligence — intelligent recovery point recommendations with blast radius analysis. Input: natural language queries, backup data snapshots, threat intelligence feeds. Output: search results with citations, anomaly alerts, recovery recommendations, data classification reports. All recovery actions require human authorization.

Technology

AI classification

Generative AISource: Generative AI: AI that generates new or synthetic content (e.g., images, videos, audio, text, code).

System name(s)

Cohesity Data Cloud (FedRAMP Moderate Authorized)

Custom-developed code

No

Public source code

Not applicable - No custom code

Sourcing

How it was built

Purchased from vendorSource: a) Purchased from a vendor

Vendor (as reported)

Cohesity, Inc.

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Yes

Privacy Impact Assessment

Not publicly available

Authorization to Operate

Yes

Demographic features

k) None of the above

Training and evaluation data

Commercial vendor-managed AI models. Gaia uses LLM models (Google Gemini or Azure OpenAI) to process queries against agency backup data. No agency data is used for model training.

Federal Data Catalog

Not applicable - Proprietary vendor training data

Governance

8 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Impact assessment in progress. Primary risks include AI-powered search exposing sensitive data to unauthorized users, false negatives in threat detection missing embedded malware in backup snapshots, and incorrect recovery point recommendations leading to data loss. Mitigated by RBAC-enforced access controls on all queries, integration with Google Threat Intelligence for malware scanning validation, human approval required for all recovery operations, and immutable audit logging. Identified through vendor FedRAMP documentation review and internal data security assessment.