Learn

Threat intelligence is analyzed, decision-ready knowledge about an adversary's capabilities, intentions, and activities, produced by collecting and evaluating information from many sources so commanders, analysts, and defenders can anticipate a threat and act before it materializes.
Last updated: November 18, 2026
Most writing on threat intelligence starts and ends with malware feeds. In defense and national security the term is older and wider: it is intelligence about who could hurt you, what they can do, and what they intend to do, built in time to change a decision. Cyber threat intelligence is one important branch of it.
Threat intelligence is analyzed, decision-ready knowledge about an adversary's capabilities, intentions, and activities, produced by collecting and evaluating information from many sources so commanders, analysts, and defenders can anticipate a threat and act before it materializes. It is also called threat intel, and in the network-defense context, cyber threat intelligence (CTI).
The word doing the work is analyzed. A list of suspicious IP addresses, a report of a ship leaving port, or a social media post is information. It becomes threat intelligence only when someone has related it to other information, judged what it means about an adversary, and put it in front of the person who has to decide. Joint doctrine draws exactly this line: information "related to other information about the OE and considered in the light of past experience" becomes intelligence, and intelligence "allows anticipation or prediction of future situations and circumstances" (JP 2-0, Joint Intelligence, Ch. I).
Key takeaways
In a defense context, a threat is the combination of a capability and the intent to use it against you. An adversary with long-range missiles and no intent to fire is a capability to watch. An adversary with intent and no means is a concern. Threat intelligence exists to assess both, and to notice when either one changes.
That framing comes straight from doctrine. JP 2-0 describes intelligence assessments that are "threat-based providing an analysis of threat capabilities and intentions," and says tactical intelligence "seeks to identify when, where, and in what strength the adversary will conduct tactical level operations." The DoD Dictionary of Military and Associated Terms defines intelligence itself as the product of collecting, processing, integrating, evaluating, analyzing, and interpreting information "concerning foreign nations, hostile or potentially hostile forces or elements."
Seen this way, cyber threat intelligence is one branch among several. The same discipline applies to a maritime militia, a terrorist cell, a hostile intelligence service, an illicit finance network, or a nation-state's missile program. What changes across branches is the data, not the method.
The government states this demand in law, not only in doctrine. The Office of the Director of National Intelligence publishes an annual unclassified threat assessment required by Section 617 of the FY21 Intelligence Authorization Act; the 2025 edition says it "reflects the collective insights of the Intelligence Community" for "policymakers, warfighters, and domestic law enforcement personnel." On the cyber side, the Cybersecurity Information Sharing Act of 2015 made the Cybersecurity and Infrastructure Security Agency the hub for exchanging threat indicators, which CISA runs through its Automated Indicator Sharing service for the "real-time exchange of machine-readable cyber threat indicators and defensive measures."
Threat intelligence is usually divided into four types. They differ by who consumes the product, the time horizon, and the questions it answers. The first three line up with the levels of intelligence in JP 2-0 (strategic, operational, tactical); the fourth, technical, is the machine-readable layer that cyber and signals work adds.
| Type | Primary consumer | Time horizon | Questions it answers | Defense example |
|---|---|---|---|---|
| Strategic | National leaders, combatant commanders, senior officials | Months to years | Which adversaries matter, what they intend, how their capabilities are trending | An assessment of a rival's long-term investment in counterspace weapons |
| Operational | Joint force and campaign planners | Weeks to months | How an adversary is likely to conduct a campaign; centers of gravity and vulnerabilities | An estimate of how an adversary would sequence a blockade |
| Tactical | Commanders, operators, defenders | Hours to days | When, where, and in what strength the threat will act; the TTPs it will use | Warning that a unit's position has been located and targeted |
| Technical | Sensors, security tools, analysts | Minutes to hours | Which specific indicators reveal the threat: signatures, selectors, hashes, domains | A file hash or emitter signature fed to detection systems |
The practical lesson is that the types are not separate products built by separate teams. A technical indicator only matters because of the tactical activity it reveals, and a tactical pattern only matters because of the strategic intent behind it. Good threat intelligence keeps those layers connected, so a commander can move from "what happened on this network" to "who is doing this and why" without losing the thread.
The threat intelligence lifecycle is the repeating process that turns a question about a threat into a finished, acted-on judgment. Industry describes it in six phases. Those phases are the joint intelligence process: JP 2-0 names "six interrelated categories of intelligence operations": planning and direction; collection; processing and exploitation; analysis and production; dissemination and integration; and evaluation and feedback.
Doctrine adds a point the six-step diagrams usually hide: the phases "occur almost simultaneously or may be bypassed." A time-critical report may go straight from processing to the commander. The lifecycle is a model of the work, not a gate sequence.
Take a single requirement: is a hostile actor preparing to target a forward logistics node? Planning turns that into specific indicators to watch. Collection pulls open-source posts mentioning the node, signals activity near it, a human report about unfamiliar vehicles, and network logs showing scans of the node's systems. Processing translates the posts, extracts the names and places in the human report, and normalizes the logs. Analysis connects the pieces: the vehicles in the human report match a group whose online accounts were discussing the node, and the network scans trace to infrastructure previously linked to the same group. The finished judgment, with its confidence level and its sources, goes to the commander, who adjusts force protection. Feedback then sharpens the next collection cycle around that group.
Notice that the cyber evidence and the physical evidence answered the same question. Treated as separate pipelines, each was a weak signal. Connected, they were warning. That is the argument for treating cyber threat intelligence as a branch of all-source work rather than a silo, and it is the same logic behind indications and warning.
Threat information is any data about a threat; threat intelligence is that data after analysis has given it meaning and context for a decision. The National Institute of Standards and Technology draws the line precisely. In NIST SP 800-150, Guide to Cyber Threat Information Sharing (2016), threat information is "any information related to a threat that might help an organization protect itself against a threat or detect the activities of an actor," while threat intelligence is "threat information that has been aggregated, transformed, analyzed, interpreted, or enriched to provide the necessary context for decision-making processes."
The distinction matters because organizations routinely hold the information without producing the intelligence.
By the numbers. Before January 6, 2021, all 10 selected federal agencies the U.S. Government Accountability Office reviewed were aware of open-source data about events planned for that day, and 7 of the 10 were aware of potential violence. Seven agencies produced 38 election-related threat products based partly on open-source data, 26 of them about January 6. Source: GAO-22-105963, Capitol Attack: Federal Agencies' Use of Open Source Data and Related Threat Products Prior to January 6, 2021 (2022). Scope: a single-event case study of selected agencies; GAO reported awareness of data and products produced, not how each was acted on.
GAO's findings are about awareness and output, not outcomes, so draw only a narrow lesson: in this case the information existed across many agencies. Holding information is not the same as producing intelligence, and whether information becomes warning depends on how well it is correlated, analyzed, and delivered, which is the back half of the lifecycle. See what is OSINT for how open-source data feeds that process.
Cyber threat intelligence is threat intelligence focused on adversaries operating in cyberspace: who is targeting your networks and systems, how they operate, and which technical indicators reveal them. It is the branch most commercial material means when it says "threat intelligence," and it has its own vocabulary, much of it set out in NIST SP 800-150:
Indicators are the fastest to use and the fastest to go stale; an adversary can change an IP address in minutes. TTPs and actor intent change slowly and are harder to hide, which is why mature programs push past indicator matching toward understanding the actor. In defense, that actor is often a state or state-sponsored group whose cyber activity is one line of effort among several, which is exactly when connecting cyber findings to the wider all-source picture pays off.
AI threat intelligence is the use of machine learning and language models to speed up the parts of the lifecycle that drown analysts in volume: processing, correlation, and cueing. AI threat detection is the narrower task of using those models to spot threat activity, such as an anomaly in network traffic, an unusual movement pattern, or a known actor's signature, in data too large for people to watch.
Applied well, AI helps at specific phases:
The constraint is that analysis and production remain a human responsibility. A model can propose that two reports describe the same group; an analyst decides whether that link holds, at what confidence, and what it means. That makes provenance non-negotiable: every AI-generated link or summary has to trace back to the source records that support it, or it cannot be briefed or defended.
The pressure for AI cuts both ways. ODNI's 2025 threat assessment warns that Beijing "is likely to feel emboldened to use malign influence more regularly in coming years, particularly as it fields AI to improve its capabilities and avoid detection," and that "Russia is using AI to create highly-capable deepfakes to spread misinformation." Adversaries using AI to scale deception raise the volume and quality of what defenders must sort through, which strengthens the case for AI on the defender's side and for keeping a traceable, human-checked line from every conclusion back to its evidence.
These terms overlap and are often used loosely. The table separates them so each can be cited on its own.
| Term | What it is | Relationship to threat intelligence |
|---|---|---|
| Threat assessment | A judgment of the level of threat. The DoD Dictionary defines it (in antiterrorism) as examining the "capabilities, intentions, and activities, past and present" of threat organizations | A product of threat intelligence, scoped to a specific threat, location, or force |
| Indications and warning (I&W) | Detecting and reporting time-sensitive signs that an adversary is preparing to act | The time-critical application of threat intelligence. See indications and warning |
| All-source intelligence | Intelligence that incorporates all available sources of information into finished products | The method threat intelligence should use; threat intelligence is all-source work focused on adversaries. See all-source intelligence |
| Cyber threat intelligence (CTI) | Threat intelligence about adversaries in cyberspace: indicators, TTPs, actors | A branch of threat intelligence, not a synonym for it |
| Threat hunting | Proactively searching one's own networks or environment for adversary activity that existing detections missed | A consumer of threat intelligence: hunters use known TTPs and indicators to decide where to look |
| Threat data or feeds | Raw or lightly processed indicators and reports | Input to threat intelligence; by NIST's definition it is threat information until analyzed |
If you are evaluating technology to support threat intelligence for a defense or national security mission, require:
Torch.AI does not sell a threat feed or produce finished intelligence. It builds reasoning infrastructure that supports the processing and analysis phases of the lifecycle, the phases where information most often stalls before it becomes warning. The infrastructure complements existing systems of record and collection systems rather than replacing them.
ORCUS is the data-availability layer: it connects to sources where they live, preserves lineage and provenance, and can run anomaly detection over a dataset query to surface outliers analysts would otherwise have to find by eye. NEXUS is the analytical layer: it extracts text and entities from unstructured reporting such as PDF and HTML documents and gives that content a semantic representation, organized into reusable knowledge domains, so it can be searched and correlated by meaning instead of keyword. HALO makes relationships understandable: it builds a mission knowledge graph, resolves when records describe the same actor or asset, and answers analyst questions with graph-grounded responses that cite their sources. Together they turn fragmented, multi-source data into coherent understanding at machine speed, with the analyst still making the call.
Ownership follows the deployment. Torch.AI's capabilities can be delivered government-owned (GOTS) when the mission requires the customer to control the reasoning layer, or commercially (COTS); which applies depends on what the customer installs and purchases. See how the products are packaged on the Torch.AI software page.
If your threat intelligence problem is that the information exists but no one can correlate it across sources fast enough to act, that is a data and reasoning problem, and it is the one this infrastructure is built for. Request a technical walkthrough and we will work through it against a representative sample of your own data, with provenance traced end to end.
Threat intelligence is knowledge about who could harm you, what they are able to do, and what they intend to do, built by analyzing information from many sources so you can act before the threat materializes. Raw data and alerts are inputs; threat intelligence is the analyzed judgment that comes out.
The four types are strategic, operational, tactical, and technical. Strategic intelligence informs senior leaders about adversary intent and long-term capability; operational intelligence informs campaign planning; tactical intelligence tells commanders and defenders when, where, and how a threat will act; and technical intelligence supplies the specific indicators that sensors and security tools detect.
The six phases are planning and direction, collection, processing and exploitation, analysis and production, dissemination and integration, and evaluation and feedback. They are the six categories of intelligence operations in Joint Publication 2-0, and in practice they often overlap or are bypassed when time is short.
Threat information is any data about a threat. Threat intelligence, as NIST SP 800-150 defines it, is threat information that has been aggregated, transformed, analyzed, interpreted, or enriched to provide the context needed for decisions. An indicator list is information; an assessment of who is behind it and what they will do next is intelligence.
AI speeds up the high-volume phases of the lifecycle: translating and extracting entities from unstructured reporting, resolving when records from different sources describe the same actor, flagging anomalies for review, and summarizing what is known with citations. Analysts still own the judgment, so every AI-generated link or summary should trace back to its source records.
AI threat detection is the use of machine learning to spot threat activity in data too large for people to monitor, such as anomalous network traffic, unusual movement patterns, or a known actor's signature. Its outputs are cues for an analyst to investigate, not finished conclusions.
No. Cyber threat intelligence is one branch. In defense and national security, threat intelligence covers any adversary capability and intent, from missile programs and maritime activity to terrorist networks and hostile intelligence services, and the strongest warning often comes from connecting cyber findings with other sources.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
Threat intelligence stalls when no one can correlate the information in time. Describe the problem, and we will scope the session. Analysts make the call. Your own data. Real work.