The Data Security Problem Holding Back Enterprise AI

AI has grown much faster than cybersecurity can catch up, and OpenAI’s Hugging Face breach is just the latest example of how good AI agents have become at discovering vulnerabilities. That leaves enterprises at risk and scrambling to figure out data access and controls. 

Data cybersecurity software company Seclore can foresee the mounting risks as enterprises consider organisation-wide AI deployment. It recently partnered with enterprise AI platform Glean to help enterprises classify and protect sensitive data as they expand AI use.

The integration connects Seclore’s ARMOR DSPM with Glean's Enterprise Graph. 

ARMOR DSPM uses data and metadata already indexed by Glean to identify sensitive information, assess exposure risks, and apply security controls—without building a separate data-discovery pipeline.

Under the partnership, ARMOR DSPM takes Glean’s existing enterprise context as an input for its classification engine. It assigns sensitivity, information type, and exposure-risk labels to files, then maps them against regulatory frameworks including GDPR, HIPAA, and India’s Digital Personal Data Protection Act.

Because ARMOR DSPM reads from Glean's existing index rather than re-crawling the same repositories, classification can cover the full scope of data already connected to Glean.

Once organisations identify sensitive information, they can apply encryption, usage controls, and access policies through Seclore's remediation tools. Classifications are written directly onto source files, so they stay with the data regardless of which application it moves through.

The same controls extend to AI-generated files. 

When an AI system draws on multiple sensitive documents to produce a report or summary, that output can itself become a data-security liability. Seclore says ARMOR can carry the source material's classification into generated files, so the output can be encrypted, access-controlled, and audited.

The integration initially targets Glean customers in regulated industries: financial services, healthcare, legal services, and government.

Company Brief

Seclore, incubated at IIT Bombay, was founded as a data security company focused on protecting information. Its products have historically centred on data-centric security and enterprise digital rights management. 

Its newer ARMOR platform combines data discovery, contextual intelligence, protection, and usage monitoring.

Glean, founded in 2019, started with enterprise search before expanding into a broader enterprise AI platform. Its Enterprise Graph connects information across an organisation's systems and maps relationships between content, people, and work. 

It now offers AI assistants and agents that use that enterprise context to answer questions and perform tasks across business applications.

Securing the Data Layer

Glean’s growth signals why securing this data layer has become urgent. 

In May, the company said it had crossed $300 million in annual recurring revenue–tripling within 15 months. Its Fortune 500 customer base had nearly doubled year-over-year, and more than 85% of customers were deploying Glean across five or more departments.

Seclore CEO Vishal Gauri believes the growth reflects a problem enterprises haven't yet solved: the pace of AI deployment is outrunning the security controls around it.

“There is a constant push and pull between the DPO and data teams, AI leadership and implementation teams, innovation groups, and cybersecurity teams,” Gauri tells AIM. 

Business teams want to deploy AI systems to improve productivity. But security teams must assess whether those systems can access personally identifiable information, intellectual property, and other sensitive enterprise data.

The problem compounds as enterprises move from AI assistants that retrieve information to agents that can act across business systems. Gauri expects data security and identity security to become more important as organisations need to set controls not just for employees, but for non-human identities as well.

“For humans, you have rights and permissions on data based on sensitivity and role. You now have to understand how to make that effective for AI agents,” he remarks.

Gauri sees partnerships with companies that hold enterprise data and context as one practical path. Rather than building security controls alongside AI platforms, vendors can integrate them into the systems enterprises already use.

“The opportunity to innovate and to partner with companies that work with enterprise data, like Glean and others, is very large,” Gauri says. “Building the trust layer directly into these systems and giving customers that comfort allows them to adopt them faster.”

The Security Question

However, enterprise-wide AI deployments are still uncommon. Organisations often confine the adoption to smaller or more targeted rollouts. 

Demand for broader deployment is growing regardless. According to McKinsey’s State of AI survey 2026, 44% of organisations report that AI is scaling across their enterprise—up from 38% in the previous year.

In such a scenario, the security question, then, is not simply whether an AI system should have access to enterprise data, but which data it can access, under what conditions and what controls should remain in place when AI models and agents use it.

The same applies to Seclore's own software. Advanced AI models can find vulnerabilities that conventional testing misses, compressing the time teams have to respond.

“We have to address them in hours and days, not weeks and months, which was the old paradigm for addressing these challenges,” Gauri says.

The deeper issue is what happens when AI systems or malicious actors get past traditional controls. Protecting the perimeter, cloud infrastructure, and applications is not enough if the data remains exposed.