OPSWAT has announced findings from “The 2026 Enterprise File Security Report: AI Is Exposing the File Security Gap.” Conducted by Osterman Research, the comprehensive survey of 254 IT and security leaders reveals a disconnect between how secure organizations believe their file lifecycles are and the effectiveness of their defences in the AI era.
AI is accelerating the volume and sophistication of cyber threats. Three out of five organisations (59%) said they expect AI to make file security more time consuming for their IT and security teams. Concern about AI-driven threats against files rose by an average of 78.9% from 12 months ago to the level anticipated 12 months from now, with fears surrounding generative AI and large language model (LLM) threats increasing the sharpest, by 105%. Furthermore, 68.1% of security leaders said that cybercriminals are more focused on circumventing their file security protections than 12 months ago.
Confidence outruns capability
Three out of four organisations (73.6%) self-assessed their organisation’s file security posture as “very” or “extremely” mature. When Osterman Research scored each respondent’s survey answers against an evidence-based maturity framework, however, no organisation (0%) gave evidence of either of the top two maturity levels, and 85.8% evidenced no more than “somewhat mature.”
What leaders believe about their security posture is unsupported by their own answers. For example, 94.5% over-assessed their maturity compared to the evidence of their file security posture. Furthermore, security leaders reported higher average confidence in detecting and preventing new and emerging threats (60.6%) than long-running threats like malware (56.6%). Against polymorphic threats that actively evade signature-based defences, extreme confidence dropped to just 12.2%.
“The takeaway from this research is the vast distance between perceived security and evidenced reality,” said Michael Sampson, Principal Analyst at Osterman Research and author of the report. “By comparing organisations’ actual capabilities against their self-ratings, we found a gap that AI-driven threats will severely test organisations over the next 12 to 24 months. You cannot defend against what you incorrectly assume is already protected.”
From detection to prevention
The financial and operational costs of this security gap are significant. In the past two years, two out of five organisations (41.7%) suffered file-related breaches or security incidents, costing an average of $2.2 million. The true breach rate is likely much higher, as 83% of those reporting no incidents admitted they lack full visibility into their file lifecycles. In other words, they couldn’t rule out a breach. Less than one in five of organisations (16.9%) claimed complete visibility into files entering, moving within, and leaving their networks.
Many still endorse detection-based approaches, but some in the industry are signaling a need for change. Of all security posture statements tested, organisations most highly rated (27.2% strong agreement) that a paradigm shift away from a “detection-first” posture is urgently required.
Content disarm and reconstruction (CDR), which neutralises threats by removing potentially malicious components from files, is already used by 51.6% of organisations, and another 8.3% plan to deploy it within 12 months, one of the highest planned adoption rates among the file security methods examined. This current increase speaks to the growing need for a prevention first methodology to better manage today’s threat landscape.
“Detection-first is reaching its limits, and this latest research shows security leaders acknowledge it,” said George Prichici, VP of Products at OPSWAT. “When AI-generated threats appear daily, a control that has to recognise a threat first will always be a step behind. Prevention-first changes the question from which threats you caught to which classes of risk you addressed.”
Where organisations failed to deploy modern file security prevention methods, the top two barriers were a lack of senior leadership support and a lack of knowledge about the methods themselves. Budget ranked third, suggesting that file security is currently more an executive-education problem than a spending problem.
Discussion about this post