Use cases covered here: insider risk management, endpoint data loss prevention, and AI usage control.
DTEX Systems was founded in 2000 in Australia by Mohan Koo, who serves as chief technology officer, with Christiaan Nillissen and Benjamin Wannan. It is headquartered in San Jose, California, and is led by chief executive Marshall Heilman. The company has raised roughly $138 million, including a $50 million Series E led by CapitalG, Alphabet's growth fund, in March 2024.
Its product, DTEX InTERCEPT, consolidates data loss prevention, user and entity behavior analytics, and user activity monitoring into one lightweight platform, with a FedRAMP-authorized variant for government. Alongside the software sits DTEX i3, an investigations and intelligence team that publishes insider threat research and works with MITRE and Five Eyes defense partners. In June 2026 the company added an AI Risk Management module covering how employees and agents use AI, along with Threat Hunter, which lets analysts run investigations in natural language.
DTEX has done more than any other vendor to make insider risk a defined category, and the i3 research is genuinely useful work that benefits the whole field. The architectural critique below is about a design choice, not about seriousness.
Insider risk is the hardest data problem because every action involved is permitted. A departing engineer cloning a repository has repository access. A sales director exporting a pipeline report has CRM access. There is no malware and no credential anomaly. The only question is whether this action, by this person, right now, is work or theft.
Two approaches have formed around that question.
The behavioral school holds that intent lives in the pattern of actions. Unusual hours, access to systems rarely touched, files compressed and renamed, collection broadening after a resignation. The sequence is the signal, and the evidence is a timeline. This is where DTEX sits, and it is the tradition UEBA and user activity monitoring come from.
The data school holds that behavior without data context cannot be read. Eleven files moved at 2am might be a backup or an exfiltration, and the difference is what is in the files. Classify the content first, then weigh behavior against it, so the risk score carries a data dimension. This is where Bold sits.
The gap between them shows up in one sentence. A behavioral product can tell you a person is risky. A data-first product can tell you what is at stake. "Risky person" is not an actionable answer, and it is not something Legal or HR can act on without a week of investigation.
The three use cases in this comparison:
Insider risk management: identifying when an employee or contractor is collecting, obfuscating, or moving data they have access to but no business reason to take, and producing evidence Legal and HR can act on.
Endpoint data loss prevention: stopping sensitive data from leaving a managed device through web uploads, email, personal cloud accounts, removable media, printing, clipboard, and screen capture.
AI usage control: seeing which AI tools, AI browsers, and coding assistants run on each device, what data goes into them, and enforcing policy without blocking the tools people need.
This is the central difference and DTEX documents its side of it plainly. InTERCEPT's data sensitivity modeling identifies file sensitivity without heavy content inspection, using digital fingerprinting and file attributes, and the company's own engineering blog describes the endpoint collector as gathering behavioral metadata with no content inspection and no local processing.
Bold classifies the content itself, on the device, across four dimensions at once: sensitive elements, business artifact, content category, and data affiliation. Source code, financial models, board material, and proprietary formats are recognized on day one with no rule or fingerprint prepared in advance. A renamed file with its extension stripped is still classified for what it contains.
That difference decides what an alert can say. One can report that a user copied 400 files to removable media at 2am. The other can report that 400 files of unreleased product design went to removable media at 2am, and can stop the copy while it is happening.
DTEX's model routes risk through detect, investigate, escalate. By the time an analyst opens the case, the data has moved.
Bold's enforcement is graduated and happens inline. It coaches when intent is unclear, redirects to a safe alternative such as a corporate AI account rather than a personal one, and blocks on clear exfiltration. About 85% of users comply when coached, which means most risky actions resolve with no incident, no ticket, and no investigation for anyone to staff. For a team that cannot grow headcount, that is the difference between a program and a backlog.
DTEX's forwarder sends metadata to the cloud, where behavioral enrichment and risk modeling happen. Bold runs small language models locally, so classification and enforcement complete on the endpoint and keep working with no connection, on untrusted networks, and inside desktop applications that encrypt their own traffic.
DTEX's answer is minimization: collect 3 to 5 MB of metadata per user per day, no screen recording, no keystroke logging, and patented pseudonymization that tokenizes usernames, email addresses, IP addresses, and device names. That is a strong answer and it is one of the reasons DTEX survives European works council reviews.
Bold's answer is locality: content is classified on the device and never transmitted for analysis, encrypted evidence goes to your own cloud tenant under your own key, Bold receives maskable metadata, and monitoring is scoped by data provenance so personal activity between personal sources and destinations can be left unmonitored entirely.
The practical question for a privacy review is not which vendor cares more. It is whether you would rather send less data to a vendor, or keep the content from leaving at all.
Behavioral baselines assume a person at the keyboard. An agent moving data has no baseline to deviate from, so there is nothing to score against. Bold reads the process tree on the device, so every action is attributed to a person or to local software acting on its own, including coding assistants operating unattended. DTEX added AI Risk Management in 2026 to cover employee and agent AI use, so ask for a demonstration of how each product attributes an action taken by an agent rather than a human.
Here is Bold's coverage by channel and operating system, including the channels where Bold monitors today rather than blocks. Ask any vendor you are evaluating, including us, for the same table.
The features below are described as DTEX documents them.
DTEX earned its position. It built the insider risk category, the i3 research is real contribution rather than content marketing, and the privacy architecture is thoughtful enough that it clears reviews other products fail. If your program is mature, if you need FedRAMP, or if your investigations depend on long-horizon behavioral timelines, DTEX is deeper than we are and we would say so.
There are four things a mature insider risk program will find missing in Bold today, and naming them is more useful than pretending otherwise: correlated multi-signal behavioral timelines, where a flight-risk indicator rising while data movement rises is itself the detection; watch lists with score multipliers for flight risk, reduction in force, and privilege misuse; chain-of-custody export built for handoff to legal; and saved, repeatable threat hunts. DTEX has that machinery and we are building it.
What we would argue is that the machinery sits on top of a signal that cannot see the data. Behavior tells you something changed. It cannot tell you what was taken, because the product deliberately does not read the content. That is a defensible design choice made for privacy and footprint reasons, and it has a cost: every serious case needs a human to reconstruct what the files were, and the reconstruction is an inference rather than a record.
Two questions worth asking both vendors. First, when an alert fires, does the product tell you what data was involved, or does an analyst work that out afterward? Second, when the evidence is handed to Legal or HR, is it a verdict with the data attached, or a set of logs someone still has to interpret? Those two answers decide how much headcount the program needs.
One thing to settle early: DTEX sponsors widely circulated insider-risk cost research, so figures from it will appear in any evaluation. Treat vendor-sponsored numbers as marketing from whichever vendor paid for them, ours included, and ask instead what the product will do on your own endpoints this quarter.
DTEX watches how people behave. Bold reads what the data is, and acts while the action is still happening.
DTEX Systems is an insider risk management company founded in 2000, headquartered in San Jose, with roughly $138 million raised including a $50 million Series E led by CapitalG. Its InTERCEPT platform consolidates DLP, UEBA, and user activity monitoring using lightweight behavioral metadata collection, and it offers a FedRAMP-authorized variant for government.
What each product reads. DTEX collects behavioral metadata and models file sensitivity without heavy content inspection. Bold classifies the content itself on the device, so an alert carries what the data was rather than only what the person did.
Not on the endpoint. DTEX's own materials describe sensitivity modeling through digital fingerprinting without heavy content inspection, and its engineering blog states the collector gathers behavioral metadata with no content inspection and no local processing. This is presented as a privacy and footprint advantage, and it is a reasonable one.
Its model is built around detection, investigation, and escalation rather than inline enforcement at the moment of action. Ask what it blocks, on which channels, and how quickly. Bold coaches, redirects, or blocks inline, with about 85% of users complying when coached.
Today, DTEX, if what you need is long-horizon behavioral correlation, watch lists with score multipliers, chain-of-custody export, and saved threat hunts. Bold does not have that machinery yet and says so.
Bold, because the classification happens at the moment of the action and travels with the event. A behavioral record tells you a sequence occurred and leaves the content to be reconstructed afterward.
They can. Some organizations run behavioral analytics for workforce-wide risk scoring and add on-device classification and prevention where the data actually moves. Bold exports its events to your SIEM, which is usually where the two meet.
It depends on whether you start from behavior or from data. For DLP and insider risk grounded in content classified on the device, with prevention at the moment of action, Bold. For file exfiltration monitoring built around employee departures, Mimecast Incydr. For organizations standardized on Microsoft 365, Microsoft Purview Insider Risk Management.