Most DLP tools catch data loss only after exfiltration

The Cloud Security Alliance has published survey findings suggesting many data loss prevention tools detect incidents only after data has left organisations. Just 16% of respondents said their most significant data loss incident was blocked before the data left.

The survey of 433 IT and security professionals found that 73% of organisations had at least one confirmed data loss incident in the past 24 months. Respondents were mainly from the United States and Canada and represented organisations ranging from fewer than 1,000 to more than 50,000 employees.

The findings point to a gap between how security teams assess newer data risks and how they enforce controls. GenAI was ranked as the top exfiltration risk by 48% of respondents, while 41% cited cloud storage uploads.

Yet strict blocking was far less common in those areas. Only 22% of organisations said they strictly block GenAI channels, and 19% said the same for cloud storage uploads.

By contrast, removable media ranked last as an exfiltration risk at 10%, but 31% of respondents said it was subject to strict blocking. That made it the most heavily restricted channel measured in the survey.

Visibility gaps

The report also found that many teams lack routine visibility into how sensitive data is sent to AI tools. While nearly half of respondents ranked GenAI as the leading risk, fewer than a third said they could produce a routine 30-day report showing sensitive data sent to those tools.

That figure was 29%. The result suggests concern over AI-related data loss is not matched by the ability to track activity consistently.

Hillary Baron, AVP of Research at the Cloud Security Alliance, said the findings raised long-standing questions about how well DLP systems were working.

"Enterprise DLP has moved beyond the question of whether it should be deployed to whether it's doing what it is intended to do. Enforcement is being softened, analyst capacity is consumed by noise, and new data-exit channels created through GenAI are exposing gaps legacy DLP was never meant to handle," Baron said.

"In the AI era, where employees and applications can move sensitive data through more channels than ever, organisations can't afford to discover gaps after the fact," she said.

Alert burden

The survey highlighted strain on security operations teams from large volumes of DLP alerts. Fifty-four per cent of respondents said they were triaging alerts where most turned out to be false positives.

That workload appears to have broader consequences. Nine in ten respondents reported at least one measurable effect from alert fatigue over the past year.

The impacts included voluntary staff departures, with 21% of respondents linking staff leaving directly to the burden created by DLP alerts.

Enforcement was also shaped by internal resistance. Seventy-one per cent of respondents said their security teams had received pushback on DLP enforcement during the past two years.

Among those organisations, 66% said they reduced enforcement for specific teams or use cases. Another 8% said they reduced enforcement across the organisation, while 3% said they reduced or paused the programme entirely.

Across all respondents, 54% said they had softened enforcement in some form. The findings suggest business disruption remains a major factor in how data controls are applied.

Jazz, which collaborated on the study, said the spread of AI tools inside companies was complicating oversight of data flows. Organisations often use far more AI services than those formally approved, it said.

"Security teams know GenAI is creating new ways for sensitive data to leave the business, but many still cannot see the full impact," said Ido Livneh, Co-founder and CEO of Jazz.

"Across the customer environments we observe, the average organisation uses nearly 170 AI tools and fewer than 10 are sanctioned. Teams need to understand how data is actually moving so they can focus on real risk without blocking legitimate work," he said.