Where does ThreatOps stand as 2026 draws to a close? The honest answer is that the practice is mature while the label is not. Intelligence tradecraft and hunting methods have a decade or more of public literature behind them, but nobody agrees on what to call the combination. This state-of-play piece pulls together the milestones, the numbers, the open arguments and the practical consequences for engineers and trainers, drawing on the timeline in the Ops Atlas from xOps School.
The label problem first
Per the Atlas, no formal standards body has defined ThreatOps. The glossary maintained by the US standards institute returns no entry, so the Atlas classifies the term as ambiguous. Vendors have used it in at least three unrelated ways: an intelligence-driven blocking service launched in 2022, the brand name of a managed detection team, and a gamified hunting-training platform that ran from 2020 to 2026. If you see the word in a job advert or a contract, ask which meaning is intended before you form a view.
The Atlas makes an editorial choice to define it as intelligence and hunting run together as one operation, and to anchor that in two working definitions: NIST’s description of threat intelligence as threat information that has been aggregated and analysed to support decisions, and the hunting definition from Splunk’s PEAK, a manual or machine-assisted process to find incidents that automated detection missed.
A fifteen-year ThreatOps timeline
The entry lays out a sequence of building blocks, and reading it in order shows how the field assembled itself.
- 2011: the kill-chain model of intrusion stages.
- 2013: the Diamond Model and the Pyramid of Pain, which ranks indicators by how costly they are for an attacker to change.
- 2015: the Sqrrl hunting model and the hunting maturity model.
- 2016: NIST SP 800-150 on information sharing.
- 2018: TaHiTI, a hunting methodology.
- 2021: STIX 2.1 and TAXII 2.1 as OASIS standards.
- 2022: the second edition of TLP handling labels, authoritative from August.
- 2023: PEAK, from Splunk’s research team.
- 2024: NIST Cybersecurity Framework 2.0, with outcomes that touch risk assessment and adverse-event analysis.
- 2026: version 1.3 of the CTI capability maturity model in January, and ATT&CK version 20 on 13 October.
What stands out is the rise in practical, procedural material since 2021: standards for exchange, labels for handling, and hunting methods that say what a finished hunt looks like. That shift matters for training, because it means an engineer can now learn the work from published procedure instead of folklore.
The newest change: a matrix that moved
ATT&CK v20 lists fifteen Enterprise tactics and 227 techniques, according to the Atlas, and introduces two new tactics, “Stealth” and “Defense Impairment”. The Atlas warns that older maps stop working. For teams that tag detections or hunt notes with technique identifiers, 2026 is a remapping year. Trainers should expect questions from students holding older course material.
What the dwell-time numbers say
The Atlas relays a March 2026 edition of a widely read industry incident report, in which the worldwide median stay of an undetected intruder grew to a fortnight from eleven days. For espionage-style cases and cases involving state-linked remote workers the figure was 122 days, and voice phishing ranked as the second most frequent initial route at eleven percent. These are the Atlas citations of a third-party report; check the original before quoting them in your own material. The direction is what matters for planning: attackers who stay unnoticed for weeks reward teams that look rather than wait.
A case from early 2024 illustrates the same point. A joint advisory by US agencies on a state-linked group said conventional indicators were largely lacking and described access kept for at least five years. A TLP:CLEAR machine-readable file was attached. From it the Atlas concludes that checking a blocklist cannot, by itself, justify saying a business was untouched.
How to read year-on-year figures sensibly
A one-year change in a median, such as fourteen days against eleven, invites headlines. Resist them. Medians depend on the mix of cases a report covers, the way incidents were discovered and which organisations contributed data. The more useful habit is the one the Atlas teaches for judgements generally: state what you believe, state how confident you are, and list the assumptions. A key assumptions check on a dwell-time claim would ask whether the sample resembles your own estate and whether the counting method changed between editions.
Three tensions the field has not settled
Build, buy or borrow. The Atlas decision tree pushes toward managed hunting when nobody owns findings, and toward a lean advisory model when no internal stakeholders want intelligence. Opinions differ on how far to take this. Expect vendors to argue for subscription breadth; the Atlas counters that feeds are alert material until measured precision proves otherwise, and that a worked quarter with fifteen percent precision should end in cancellation, not expansion.
Coverage versus value. Heat maps are still popular. The Atlas calls them a vanity measure, notes that ATT&CK itself says tactics are not ordered, and cites PEAK on the point that plenty of techniques were never meant to be detections.
Sharing versus caution. Voluntary sharing under Article 29 of NIS2 is available in Europe, while the US sharing protections from 2015 run to 11 December 2026, so check the status. Treat both as legal topics to confirm with counsel.
Controls and compliance: optional by design
One more data point for planners. The Atlas notes that NIST SP 800-53 revision 5 carries a hunting control, RA-10, which sits in none of the baselines and is optional, alongside a programme-management control on threat awareness. Nobody is forced into hunting by that catalogue. The pressure comes from evidence about slow discovery, not from audit checklists, and trainers should say so plainly to students who hope for a compliance shortcut.
What this means for engineers
The practical consequences are modest and concrete. Put an expiry on every indicator you store, because the exchange format leaves it optional. Rate every source. Keep blocking behind measured precision. Schedule hunts about every two weeks, cover one or two behaviours, and demand an artefact at the end. Report ratios, not volumes. None of it needs exotic tooling; the Atlas notes that a version-controlled repository is a legitimate starting point, with open-source intelligence platforms available once volume grows.
If your day job is in delivery, the skills overlap with analytical habits we discuss elsewhere. A separate article on sharpening analytical judgement when troubleshooting DevOps issues covers the reasoning habits that make hunts better, and the article on a skills roadmap for IT transformation helps you place ThreatOps in a wider learning plan.
What this means for learners and trainers
Curriculum design is where the label problem bites hardest. A course called ThreatOps might be tool training, a hunting workshop or an intelligence analysis class. Ask for the syllabus and look for the elements the Atlas treats as core: question writing, source grading, indicator lifecycle, hunt design with artefacts, handling labels and outcome metrics. If a syllabus is mostly a tour of products, it is a product course regardless of title.
For self-study, the Atlas gives a thirty-sixty-ninety outline that works as a lab plan: owner and five questions first, two hunts and sharing rules second, a platform decision and measurable ratios third. The maturity ladder (absent, consuming, requirements-driven, hunting programme, intelligence-led) is aligned with FIRST stages and the capability models, so it doubles as a map of competencies.
Predictions we are comfortable making
- Terminology will stay fuzzy through 2027; practice will continue to converge on procedure.
- Remapping to the new ATT&CK structure will occupy many teams for a quarter or two.
- Pressure to justify subscriptions with measured precision will grow as budgets tighten.
- Hunt notes that a stranger can rerun will become a quality marker in audits.
These are judgements, not facts, and the Atlas itself would remind us to attach a confidence level to them rather than a bare probability word. We would hold them with moderate confidence.
For the full detail behind every item in this article, read the Atlas ThreatOps entry and check the primary sources it cites before you build anything on a number.

