Query Public AI Services: recognize the technique

Lesson Purpose

What you are learning

Query Public AI Services means using a public AI service, such as a large language model, to help collect, combine, summarize, or analyze information during reconnaissance. An adversary might ask an AI service about an organization's people, technologies, business relationships, public contact details, or other information that is already available online. The important idea is not that the AI service creates secret facts. It can make public information easier to organize and connect, which may help someone prepare later targeting or social engineering.

Learning Objectives

Describe Query Public AI Services in ATT&CK terms.
Explain how public AI services can support reconnaissance by organizing public information.
Name at least one defensive observation or control relevant to the behavior.

Why This Matters

Public AI services can quickly summarize information that would otherwise require many separate searches. This can make ordinary public details more useful when they are combined into a single picture of an organization. Defenders should therefore review what information is publicly available, consider how separate facts may be combined, and use normal verification controls so employees do not trust a request only because it contains accurate background details.

OWASP Web Security Testing Guide (WSTG)

Use a repeatable web-testing method

Connect this ATT&CK training activity to the WSTG approach: establish expected behavior, identify the relevant attack surface, change one controlled input or observation at a time, compare the response with the baseline, and record evidence before drawing a conclusion.

For reconnaissance-oriented lessons such as T1595, treat reachability, exposed services, and observable application behavior as information gathering. A reachable service is evidence about exposure; it is not by itself proof of a vulnerability.

Open WSTG web security fundamentals

Exercise

Before You Practice

Practice Preparation

Use these lesson concepts for the ungraded practice activity below. You do not need outside research.

Public AI services can summarize or connect public information about people, technologies, business relationships, contact details, projects, or other published facts.
Defenders can review public websites, search results, approved exposure inventories, and other organization-owned public content to understand what can be combined.
Reduce risk through data minimization/publication review and by training personnel to verify information-seeking requests rather than trusting apparently well-informed messages.

This is ungraded practice. Use the lesson material above; exact wording is not required. Using only the Lesson Purpose, What You Are Learning, Why This Matters, and Defender Perspective sections, identify three kinds of public information an AI service could help summarize or connect. Then name one defender-side source that could help review the organization's public exposure and two controls that could reduce unnecessary disclosure or social-engineering risk. No external target interaction is required, and do not query a public AI service about a real person or organization for this exercise.

Knowledge Check

Check your understanding

Answer the questions shown below. You must answer at least 2 correctly before marking this concept lesson complete.

1. What best describes Query Public AI Services? Using a public AI service to help combine, summarize, or analyze information for reconnaissance Taking control of an existing employee account Scanning an authorized lab host for open network ports

2. Why can public AI services make reconnaissance more useful? They can organize and connect public facts that are spread across different sources They automatically prove that every public fact is current and accurate They give an attacker direct access to private organizational systems

3. Which information could a public AI service help summarize during reconnaissance? Public details about people, technologies, contacts, or business relationships Only data stored inside an isolated system that has never been published Only network packets captured from the defender's private internal network

4. Why may direct detection of this technique be difficult? An adversary can query a third-party public AI service outside the organization's systems Public AI services never produce any logs anywhere The technique always requires malware to run on an employee workstation

5. Which defensive action best reduces risk from this technique? Review public information, remove unnecessary exposure, and independently verify sensitive requests Publish more internal details so outside information is easier to validate Assume any request containing correct public facts must be legitimate

6. Which conclusion best reflects how this information should be interpreted and managed? AI can make public information easier to combine, so defenders should manage exposure and verify context Using public AI services automatically means an organization has been compromised Public information is harmless whenever each individual fact seems low risk

Learning Moment

Public AI services can make existing public information easier to combine and use. Defenders should reduce unnecessary exposure and verify sensitive requests independently.

Defender Perspective

Directly detecting an adversary's query to a third-party public AI service may be difficult because the activity can occur outside the organization's systems. Useful evidence can instead include public-content inventories, website and document reviews, exposed-contact information, code-repository findings, brand-monitoring reports, web-proxy or DNS logs for the organization's own AI-service use, data-loss-prevention alerts, and reports of suspicious messages that use accurate organizational details. Investigators should ask whether the information was intentionally public, whether it is still needed, whether several public facts create additional risk when combined, and whether a later message or request relied on those details. Risk reduction includes publication review, data minimization, secret scanning, removing obsolete public material, clear rules for entering sensitive data into public AI services, phishing-resistant authentication, and independent verification of sensitive requests.

Continue Your Learning Pathway

Next Step · Continue ATT&CK Training

Continue building your reconnaissance and resource-development foundation with the next ATT&CK lesson.

Complete this lesson first. Finish the required lesson activity and knowledge check, if present, before continuing.

Return to dashboard