# triager-canary-ind-vk44 Market Research Report - Global

**Generated on:** 2026-08-22 08:15:00.316678  
**Industry:** triager-canary-ind-vk44  
**Geography:** Global  
**Details:** triager cross-session probe mk99

---

# Opaque Label, Actionable Global Cyber-Deception Market Map

## Executive Summary

- **Scope Confidence Is Low**: Neither `triager-canary-ind-vk44` nor `triager cross-session probe mk99` resolves to a publicly documented industry, product, standard, company, or benchmark -> do not attach a market size, competitor list, or investment thesis to either identifier until the owner supplies a canonical definition.
- **Best-Fit Adjacent Market**: The most coherent public interpretation is the global cyber-deception and canary-technology market: decoy systems, honey credentials, canary tokens, identity lures, and adversary-engagement platforms. NIST treats deception as one of several techniques within a broader cyber-resiliency system, not as a complete security architecture [21] -> use this report as a conditional market map, not as proof that VK44 or MK99 belongs to this category.
- **Market Growth Is Real but Measurement Is Unstable**: Published estimates put the adjacent 2026 market at **$2.24B to $2.7B**. Forecasts range from **5.2% to 16.9% CAGR**, with endpoints from more than **$3.12B in 2032** to **$7.9B in 2033** [18][19][2] -> plan with low, base, and high cases instead of one headline forecast.
- **Identity and Cloud Are Replacing Standalone Honeypots**: SentinelOne, Proofpoint, and Commvault absorbed Attivo Networks, Illusive, and TrapX, respectively, while Tracebit and Acalvio position deception around cloud identities, credentials, and workloads [12][5][7][11] -> evaluate platforms by control-plane coverage and integration, not by decoy count alone.
- **Demand Is Supported by Breach Economics**: IBM reports a **$4.99M** global average breach cost and a **12%** year-over-year increase, while Verizon reports that **31%** of breaches now begin with software vulnerabilities [8][6] -> prioritize deception where it can expose post-compromise movement after preventive controls fail.
- **Low-Noise Detection Is the Core Value Proposition**: Thinkst says its canaries can be deployed in under two minutes, while Docker reports a notably low false-positive rate and reduced noise from its Tracebit deployment [14][13] -> pilot against alert precision, time to investigation, and coverage drift rather than accepting marketing claims.
- **Adoption Barriers Are Organizational as Well as Technical**: SANS identifies lack of awareness, perceived complexity, unclear ROI, integration challenges, and legal uncertainty as real-world concerns limiting adoption [17] -> require a deployment owner, an incident playbook, integration testing, and an evidence-based business case before scaling.
- **MK99 Must Remain Outside the Forecast**: No public evidence establishes what a "cross-session probe" means here, whether MK99 is a version, test fixture, security control, or internal experiment -> treat MK99-specific TAM, pricing, installed base, performance, and competitors as **not publicly determinable**.

## Scope Gate: VK44 and MK99 Are Not Public Market Definitions

The supplied industry label appears to be an opaque identifier rather than a recognized market taxonomy. Searches of the exact strings and their plausible components did not find an authoritative definition, vendor page, standards entry, analyst category, patent, or public product record. Component searches produced unrelated results spanning industrial data historians, hospitality software, methane sensors, staged software releases, cyber deception, and Windows cross-session activation. That dispersion is evidence of ambiguity, not evidence that all those categories belong in one market.

This report therefore applies a two-layer scope:

1. **Primary conclusion**: a defensible literal market report for `triager-canary-ind-vk44` or `triager cross-session probe mk99` cannot be produced from public evidence.
2. **Conditional adjacent analysis**: if "canary," "triager," and "probe" refer to security tripwires that detect and help investigate intrusions across sessions, the closest established category is global cyber-deception technology.

That interpretation aligns with authoritative frameworks. NIST defines cyber resiliency as the ability to "anticipate, withstand, recover from, and adapt to" adverse conditions or compromise [21]. Its framework includes goals, objectives, techniques, implementation approaches, and design principles [21]. MITRE Engage organizes adversary engagement through goal, approach, and activity across prepare, operate, and understand phases [1]. Neither framework identifies VK44 or MK99.

The distinction matters commercially. Market size, growth, segmentation, and competitive share can be discussed for cyber deception. They cannot be assigned specifically to MK99. A valid next input would include the identifier's owner, product description, buyer, use case, deployment environment, commercial model, and relationship to recognized categories. Without those fields, any MK99 revenue estimate would be fabricated.

**Decision-ready insight:** use the adjacent report for category exploration and vendor screening, but place a hard scope gate before investment, procurement, or product strategy. The gate should require a one-sentence canonical definition such as: "MK99 is a cloud-security control that deploys cross-session decoy credentials and triages their activation."

## A $2.24B-$2.7B Adjacent Market With Wide Forecast Dispersion

The public forecasts agree on growth but not on magnitude. This is common in emerging security categories because some analysts count only dedicated deception platforms, while others include identity threat detection and response, honeypots, breach detection, threat intelligence, or features bundled into larger security suites.

| Source and scope label | Base estimate | Forecast | Implied message |
|---|---:|---:|---|
| Mordor Intelligence, cyber deception | **$2.24B in 2026** | **$4.12B by 2031**, **13.01% CAGR** [19] | Moderate expansion from a defined cyber-deception base |
| Grand View Research, deception technology | **$2.4B in 2025**, **$2.7B in 2026** | **$7.9B by 2033**, **16.9% CAGR** [2] | Broader or faster-growing category definition |
| Polaris Market Research, deception technology | Base value not available in the evidence reviewed | More than **$3.12B by 2032**, **5.2% CAGR** [18] | Much more conservative trajectory |
| Grand View regional and vertical split | 2025 data | North America **34.4%**; BFSI **22.4%** [2] | Early spending concentrates in mature security markets and regulated buyers |

The takeaway is not that one forecast must be correct. The **11.7 percentage-point gap** between the lowest and highest published CAGR is itself a strategic signal: category boundaries remain fluid. A five-year plan should model a conservative case near 5%, a base case near 13%, and an upside case near 17%, but should not assign probabilities until the product definition is settled.

Demand-side evidence supports continued spending. IBM's 2026 report places the global average cost of a breach at **$4.99M**, up **12%**, and reports a **56%** increase in AI-driven attacks [8]. Verizon reports vulnerabilities as the initial route in **31%** of breaches [6]. Deception does not prevent those entry events; its economic role is to create high-signal opportunities to detect an attacker who begins reconnaissance, steals credentials, escalates privileges, or moves laterally afterward.

Broader market context also favors security-enabling infrastructure. Forrester identifies infrastructure, data and AI, and cybersecurity and identity as the three technology-market groups broadly positioned for growth as enterprises scale AI [25]. This supports the direction of travel, but it does not validate any specific deception-market forecast.

**Decision-ready insight:** size the opportunity from the bottom up. Count addressable protected identities, cloud accounts, workloads, branches, and high-value data domains; apply realistic price and attach-rate assumptions; then use analyst forecasts only as an external reasonableness check.

## Deception Evolves From Static Honeypots to Identity and Cloud Controls

Traditional honeypots emulate vulnerable hosts or services. Modern deception platforms distribute fake credentials, documents, service accounts, cloud resources, identity paths, and decoy applications throughout real environments. The mechanism is attractive because legitimate users normally have little reason to touch a decoy, so interaction can produce a stronger signal than generic anomaly detection.

NIST provides the theoretical foundation. It identifies five desired effects on adversaries: **redirect, preclude, impede, limit, and expose**. Deception contributes through actions such as deter, divert, deceive, detect, reveal, and scrutinize [21]. NIST also notes that a sandbox or deception environment can be created dynamically in response to suspicious behavior and subsequent activity diverted there [21].

MITRE Engage turns that concept into an operational process rather than a product checklist. It says internal deception infrastructure can expose, manipulate, and help defenders understand adversaries [1]. In nine MITRE elicitation operations, initial indicators averaged about two per operation; after adversary engagement, MITRE reports collecting an average of **40 new pieces of intelligence per operation** [1]. The result is useful evidence for the mechanism, although it should not be generalized into a guaranteed commercial outcome.

Three technology shifts define the market:

| Shift | Mechanism | Commercial implication |
|---|---|---|
| Network to identity | Fake credentials, service accounts, privilege paths, and directory artifacts expose credential theft and lateral movement | ITDR and XDR vendors can bundle deception with endpoint and identity telemetry |
| Appliance to cloud-native | Decoy buckets, secrets, identities, APIs, and Kubernetes resources follow ephemeral infrastructure | Buyers favor infrastructure-as-code deployment and multi-account coverage |
| Alerting to adversary intelligence | Higher-interaction environments observe tools, techniques, and objectives | Sophisticated buyers may value intelligence quality, not only detection count |
| Static to adaptive | Platforms update or generate deceptive assets as environments change | Automation reduces drift but introduces governance and validation requirements |

The table shows why the category is converging with identity, cloud security, and threat intelligence. It also explains forecast disagreement: revenue may be recorded under deception, ITDR, XDR, cloud detection, or data resilience depending on the vendor and analyst.

**Decision-ready insight:** for a putative MK99 product, define which layer it occupies. A cross-session probe might be a token, identity lure, telemetry collector, correlation feature, or full adversary-engagement system. Each has a different buyer, competitive set, price basis, and market size.

## Seven Major Players and a 2022 Consolidation Inflection

The market combines independent specialists with deception capabilities embedded in broader platforms. Structured company data identifies Thinkst, CounterCraft, Tracebit, TrapX, Illusive, and Attivo as notable category participants, with several specialists now acquired [26].

| Player or product | Positioning and mechanism | Ownership or scale signal | Best-fit buyer |
|---|---|---|---|
| **Thinkst Canary** | Hardware, virtual, cloud, and container canaries plus tokens such as fake AWS keys and documents [13] | Independent, South Africa-founded specialist; public package lists **$7,500 per year for five canaries** and unlimited tokens [26][13] | Teams seeking simple, low-overhead tripwires |
| **Tracebit** | Cloud-native AWS and identity canaries, credentials, and safe decoy resources | UK Series A company with **$25M** disclosed funding [26] | Cloud-first engineering and security teams |
| **Acalvio ShadowPlex** | AI-positioned, agentless deception across cloud workloads and IAM, including AWS, Azure, and GCP | US Series A specialist; its public-sector material targets living-off-the-land, zero-day, APT, and ransomware use cases [7] | Large enterprises and public-sector environments |
| **CounterCraft** | Deception environments designed to produce threat intelligence and observe attacker behavior | US Series A company with **$11.44M** disclosed funding in the structured record [26] | Governments and intelligence-led security teams |
| **SentinelOne Singularity Identity** | Identity deception combined with automated disruption and XDR | SentinelOne agreed to acquire Attivo Networks for **$616.5M** in cash and stock [5] | Existing SentinelOne customers consolidating endpoint and identity controls |
| **Proofpoint Identity Threat Defense** | Agentless identity-vulnerability discovery, deception-based detection, and forensic collection | Proofpoint completed its acquisition of Illusive on **December 28, 2022** [11] | Identity-centric enterprises and Proofpoint accounts |
| **Commvault ThreatWise** | Threat sensors and cyber deception integrated with data protection and recovery | Commvault acquired TrapX in **February 2022** and subsequently integrated its deception technology [12] | Buyers linking early detection with cyber recovery |

Other visible competitors include Fidelis, CyberTrap, Fortinet, RevBits, Rapid7, and open-source honeypot projects. Their inclusion varies by analyst taxonomy, reinforcing the need to define the purchasing use case before constructing share estimates.

The 2022 acquisitions mark the strategic inflection. Attivo moved into an autonomous-security and identity platform; Illusive moved into Proofpoint's identity defense; TrapX moved into Commvault's resilience stack. The mechanism is distribution: a specialist feature can reach more customers when attached to endpoint agents, identity telemetry, or protected data. The trade-off is reduced category transparency because deception revenue becomes bundled and no longer appears as a standalone line item.

**Decision-ready insight:** shortlist both specialists and suite vendors. Specialists may offer deeper deception workflows; suite vendors may lower integration and procurement costs. Score the actual deployment architecture rather than equating company size with product fit.

## Case Studies: Simplicity, Cloud Scale, and Platform Consolidation

### Thinkst turns silence into the product

Thinkst's strategy is deliberate simplicity. It deploys systems that resemble file servers, routers, web servers, or cloud assets and alerts when someone investigates them. The hosted console supports email, text, Slack, webhook, and Syslog notifications [13]. Thinkst claims setup in under two minutes and says even customers with hundreds of canaries receive only a handful of events per year [13]. These are vendor-reported claims, not independent benchmarks.

The economic model is unusually transparent for enterprise security: **$7,500 annually for five canaries**, unlimited Canarytokens, a private token server, hosted console, and support [13]. This makes Thinkst a useful pilot benchmark. It reveals that buyers may value operational quiet and fast deployment more than advanced analytics. The risk is that simple decoys can become recognizable or fail to match a changing environment; Thinkst itself notes a DNS dependency for console communication [13].

### Docker uses Tracebit for cloud-native defense in depth

Docker's case is more representative of the emerging cloud model. Tracebit reports deploying AWS canaries, canary credentials, and Okta canaries, with Terraform deployment to an AWS account in as little as **10 minutes** [14]. It integrated alerts into Docker's existing workflows through Panther and connected with deployment pipelines and SIEM processes [14].

Docker reports a notably low false-positive rate, reduced noise, and improved detection and response for data exfiltration, lateral movement, privilege escalation, supply-chain attacks, and credential theft [14]. The case demonstrates the mechanism that cloud buyers want: canaries deployed as code and routed into existing SecOps systems. However, it is hosted by the vendor, gives no controlled comparison or ROI figure, and should therefore be validated in a buyer's own environment.

### Acquisitions convert a category into platform features

SentinelOne's **$616.5M** Attivo transaction is the clearest valuation signal in the sector [5]. Proofpoint's Illusive purchase and Commvault's TrapX purchase in the same year show three different strategic destinations: XDR/identity, human-centric and identity security, and data resilience [12][11].

The divergence is instructive. The same foundational mechanism, deceptive assets that expose attackers, can support three budgets and buying centers. A new entrant should not position itself merely as "better deception." It should choose whether its economic buyer owns identity, SOC operations, cloud infrastructure, or cyber recovery.

**Decision-ready insight:** the strongest go-to-market design starts with one operational wedge. For an internal MK99 concept, a credible wedge could be cross-session identity misuse detection, but only if "cross-session" is formally defined and tested.

## Buyer Economics and an Evidence-Based Deployment Scorecard

Deception has a plausible cost argument: it seeks to improve signal quality and detect activity after a breach while avoiding the volume of alerts generated by broad behavioral monitoring. But public ROI evidence remains weak. SANS specifically lists unclear ROI, perceived complexity, integration challenges, and legal uncertainty among adoption concerns [17]. Buyers should therefore treat a pilot as an instrumentation exercise, not merely a proof that alerts can fire.

The following scorecard is an analyst-designed procurement framework. It does not claim universal thresholds; each buyer should baseline current performance and set pass/fail targets before deployment.

| Evaluation dimension | Metric | Pilot question | Warning sign |
|---|---|---|---|
| Coverage | Relevant identities, accounts, workloads, and segments carrying credible decoys | Does placement cover likely attacker paths? | High raw decoy count but gaps around critical assets |
| Signal quality | Confirmed unauthorized triggers divided by all triggers | Are alerts actionable without extensive triage? | Routine scanners or legitimate tools trigger repeated noise |
| Detection speed | Change in time from malicious interaction to SOC acknowledgment | Does the control materially improve MTTD? | Alerts arrive quickly but wait in an unowned queue |
| Investigation value | New indicators, behaviors, and affected assets discovered per event | Does the platform add context beyond the existing SIEM or EDR? | Duplicate alerts with no new evidence |
| Realism | Success against internal red-team fingerprinting and bypass tests | Can realistic attackers distinguish decoys? | Static names, services, or credentials reveal the control |
| Safety | Isolation, egress limits, data handling, and kill controls | Can a high-interaction decoy be misused? | Decoys can reach production or external targets |
| Integration | Tested SIEM, SOAR, ticketing, identity, cloud, and response workflows | Does an alert execute a documented playbook? | Proprietary console becomes another unattended pane |
| Operating cost | Deployment, tuning, refresh, investigation, and training hours | Is total cost lower than the detection value created? | Decoys drift as infrastructure changes |
| Commercial fit | Annual cost per protected critical asset or identity | Does the price scale with the buyer's architecture? | Pricing scales with decoy volume rather than risk reduction |

The scorecard shifts evaluation away from demo theatrics. A successful pilot should demonstrate realistic placement, high-quality alerts, useful context, controlled containment, and an owned response path. Thinkst's public package can provide a transparent low-complexity cost reference, but cloud, identity, and enterprise platforms will usually require quotes.

**Decision-ready insight:** require vendors to run the same red-team scenario and report the same metrics. Do not compare one vendor's alert count with another vendor's "AI detections" unless the test environment, attack path, and response workflow are identical.

## Risks: Category Ambiguity, Evasion, Safety, and Unclear ROI

**Category risk is the first-order issue.** Forecasts differ because deception overlaps with ITDR, XDR, cloud security, threat intelligence, and resilience. An opaque label such as VK44 magnifies this problem. A product can appear to address a multibillion-dollar market while actually selling a narrow feature into a much smaller budget pool.

**Fingerprinting and evasion remain technical risks.** A decoy that does not resemble the surrounding environment may be avoided. More interactive systems can collect richer telemetry but create more containment and maintenance work. NIST's broader warning applies: no single cyber-resiliency technique or set of approaches is universally optimal, and organizations need not employ every technique to meet stakeholder objectives [21].

**Operational ownership can fail even when detection works.** Alerts need routing, severity logic, evidence retention, and rehearsed response. Deceptive credentials and documents must be governed so that legitimate automation does not touch them. Cloud decoys must keep pace with account creation, region changes, and infrastructure-as-code updates. SANS's finding that deception remains underused despite its theoretical value suggests that implementation and organizational friction are material [17].

**Legal and ethical boundaries matter.** Passive decoys inside an owned environment generally present a different risk profile from active engagement that manipulates an external actor. Data collection, privacy, employee monitoring, egress, and cross-border telemetry require review. The more a platform moves from detection toward engagement, the more important legal authorization and containment become.

**Evidence quality is uneven.** Docker provides a named customer and operational detail, but the case remains vendor-hosted. Thinkst provides public pricing and concrete deployment claims, but not an independent controlled trial. Analyst forecasts disagree sharply. Acquisition prices reveal strategic value but do not establish current standalone revenue.

**Control substitution is dangerous.** Deception assumes that attackers can bypass preventive controls, but it does not replace vulnerability management, identity hardening, EDR, logging, backups, or incident response. NIST explicitly states that cyber-resilient systems cannot defend against all hazards at all times [21].

**Decision-ready insight:** place deception inside defense in depth. Approve scale only after a pilot validates realism, workflow integration, containment, maintenance burden, and incremental detection value.

## Synthesis

The adjacent market contains four distinct strategies that should not be collapsed into one feature checklist.

| Strategy | Mechanism | Scope | Main trade-off | Evidence base | Time horizon |
|---|---|---|---|---|---|
| Thinkst-style tripwires | Simple decoy services, devices, documents, and credentials | Network, endpoint-adjacent, and multi-cloud placements | Fast and quiet, but potentially less adaptive | Transparent pricing and concrete vendor claims [13] | Immediate pilot and operational use |
| Tracebit-style cloud deception | Infrastructure-as-code canaries across cloud and identity systems | AWS and identity coverage demonstrated in the Docker case | Cloud-native scale, but customer evidence remains vendor-hosted | Named deployment, workflow integration, and reported low noise [14] | Near-term cloud security expansion |
| Acalvio and CounterCraft-style platforms | Dynamic decoys, adversary observation, and threat intelligence | Enterprise, public sector, cloud, IAM, and intelligence operations | Rich context, but higher complexity and harder ROI attribution | Product claims, case libraries, and framework alignment [26] | Multi-year strategic capability |
| SentinelOne, Proofpoint, and Commvault suites | Deception integrated into identity, XDR, or resilience platforms | Existing endpoint, identity, and protected-data estates | Easier distribution and integration, but less pricing and category transparency | Three 2022 acquisitions, including Attivo at $616.5M [12][5][11] | Ongoing platform consolidation |
| NIST and MITRE framework-led operations | Goal-driven cyber resiliency and adversary engagement | Architecture and operating model rather than a single tool | Strong conceptual grounding, but requires process maturity | NIST framework and MITRE's nine-operation evidence [21][1] | Long-term program design |

Three tensions emerge. First, **simplicity versus intelligence**: a quiet token may be easier to operate, while a high-interaction environment can reveal more about an adversary. Second, **specialist depth versus suite economics**: specialists can innovate around realism and placement, while platforms can bundle deception into identity, endpoint, or recovery contracts. Third, **detection versus engagement**: passive tripwires fit conventional SOC workflows, while active adversary engagement demands stronger governance.

For `triager-canary-ind-vk44`, the correct strategic sequence is therefore:

1. Resolve the taxonomy and define MK99.
2. Identify the buyer and protected object: session, identity, workload, network, or data.
3. Select the relevant comparison set from the four strategies above.
4. Run a controlled pilot using the deployment scorecard.
5. Construct a bottom-up market model only after price basis and addressable units are known.

The adjacent cyber-deception market is attractive: it has multibillion-dollar estimates, double-digit base-case growth, active specialists, suite consolidation, and clear breach-driven demand. The supplied identifier, however, is not yet a market. The highest-value action is not to force a forecast onto MK99; it is to convert the opaque label into a testable product definition.

## References

1. *MITRE Engage™ | An Adversary Engagement Framework ...*. https://engage.mitre.org/
2. *Deception Technology Market (2026 - 2033)*. https://www.grandviewresearch.com/industry-analysis/deception-technology-market-report
3. *7 top deception technology vendors for active defense*. https://www.techtarget.com/cybersecurity/tip/7-top-deception-technology-vendors-for-active-defense
4. *A comprehensive survey on cyber deception techniques to ...*. https://www.sciencedirect.com/science/article/pii/S0167404824000932
5. *SentinelOne to Acquire Attivo Networks, Bringing Identity ...*. https://www.sentinelone.com/press/sentinelone-to-acquire-attivo-networks-bringing-identity-to-xdr/
6. *2026 Data Breach Investigations Report (DBIR)*. https://www.verizon.com/business/resources/reports/dbir/
7. *Acalvio ShadowPlex for Public Sector Organizations*. https://static.carahsoft.com/concrete/files/4217/3075/1001/ShadowPlex_for_PS_Orgs_wrapped2.pdf
8. *Cost of a Data Breach Report 2026*. http://ibm.com/reports/data-breach
9. *Cyber Deception | Explore*. https://www.commvault.com/explore/cyber-deception
10. *CounterCraft: Deception-Powered Threat Intelligence*. https://www.countercraftsec.com/
11. *Proofpoint Closes Acquisition of Illusive*. https://www.proofpoint.com/us/blog/corporate-news/proofpoint-closes-acquisition-illusive-1
12. *Welcoming TrapX to the Commvault Family*. http://commvault.com/blogs/welcoming-trapx-to-the-commvault-family
13. *Thinkst Canary*. http://canary.tools/
14. *Docker | Tracebit Customer Stories*. https://tracebit.com/customer/docker
15. *Tracebit | The answer to Assume Breach*. http://tracebit.com/
16. *SP 800-160 Vol. 2 Rev. 1, Developing Cyber-Resilient ...*. https://csrc.nist.gov/pubs/sp/800/160/v2/r1/final
17. *Addressing Barriers in the Adoption of Cyber Deception ...*. https://www.sans.org/white-papers/breaking-through-deception-addressing-barriers-adoption-cyber-deception-technologies
18. *Deception Technology Market Size $3.12 Billion by 2032 ...*. https://www.polarismarketresearch.com/press-releases/deception-technology-market
19. *Cyber Deception Market - Size, Share & Industry Analysis*. https://www.mordorintelligence.com/industry-reports/cyber-deception-market
20. *Acalvio Customer Reference: Research University*. https://www.acalvio.com/resources/case-studies/acalvio-shadowplex-customer-reference-research-university/
21. *Developing Cyber-Resilient Systems; A Systems Security Engineering Approach*. https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-160v2r1.pdf
22. *Identity Threat Detection & Response Solutions | Proofpoint US*. https://www.proofpoint.com/us/products/identity-threat-detection-response
23. *Singularity™ Identity Security Platform | SentinelOne*. https://www.sentinelone.com/platform/identity/
24. *Threatwise | Commvault *. https://www.commvault.com/platform/threatwise
25. *http://investor.forrester.com/news-releases/news-release-details/forrester-introduces-ai-disruption-model-assess-ais-impact*. http://investor.forrester.com/news-releases/news-release-details/forrester-introduces-ai-disruption-model-assess-ais-impact
26. *Tracxn Private Data*. https://tracxn.com