The 2025 State of AI Data Safety Report reveals a widening contradiction in enterprise security: AI adoption is sort of common, but oversight stays restricted. Eighty-three p.c of organizations already use AI in day by day operations, however solely 13 p.c say they’ve robust visibility into how these programs deal with delicate knowledge.
Produced by Cybersecurity Insiders with analysis help from Cyera Analysis Labs, the research displays responses from 921 cybersecurity and IT professionals throughout industries and group sizes.
The info exhibits AI more and more behaving as an ungoverned id — a non-human person that reads sooner, accesses extra, and operates repeatedly. But most organizations nonetheless use human-centric id fashions that break down at machine pace. Consequently, two-thirds have caught AI instruments over-accessing delicate info, and 23 p.c admit they don’t have any controls for prompts or outputs.
Autonomous AI brokers stand out as essentially the most uncovered frontier. Seventy-six p.c of respondents say these brokers are the toughest programs to safe, whereas 57 p.c lack the power to dam dangerous AI actions in actual time. Visibility stays skinny: practically half report no visibility into AI utilization and one other third say they’ve solely minimal perception — leaving most enterprises not sure the place AI is working or what knowledge it touches.
Governance constructions lag behind adoption as nicely. Solely 7 p.c of organizations have a devoted AI governance staff, and simply 11 p.c really feel ready to fulfill rising regulatory necessities, underscoring how shortly readiness gaps are widening.
The report requires a shift towards data-centric AI oversight with steady discovery of AI use, real-time monitoring of prompts and outputs, and id insurance policies that deal with AI as a definite actor with narrowly scoped entry pushed by knowledge sensitivity.
“AI is now not simply one other instrument — it’s appearing as a brand new id contained in the enterprise, one which by no means sleeps and sometimes ignores boundaries,” mentioned Holger Schulze with Cybersecurity Insiders. “With out visibility and strong governance, enterprises will preserve discovering their knowledge in locations it was by no means meant to be.”
Because the report cautions: “You can not safe an AI agent you don’t determine, and you can not govern what you can not see.”
The total 2025 State of AI Data Safety Report is accessible for obtain at: https://www.cybersecurity-insiders.com/portfolio/2025-state-of-ai-data-security-report-cyera/
Media Contact: Andrea.Chilkott@cybersecurity-insiders.com
About Cybersecurity Insiders
Cybersecurity Insiders supplies strategic perception for security leaders, grounded in additional than a decade of impartial analysis and trusted by a world group of 600,000 cybersecurity professionals. We translate shifting market developments into clear, actionable steerage that helps CISOs strengthen their applications, make knowledgeable expertise choices, and anticipate rising dangers.
We join practitioners and innovators by giving CISOs the readability wanted to navigate a loud market whereas serving to answer suppliers align with real-world priorities. We drive this alignment by way of evidence-backed analysis, strategic CISO guides, impartial product opinions, data-driven message validation, and peer-validated recognition by way of the Cybersecurity Excellence Awards and AI Chief Awards.
Extra: https://cybersecurity-insiders.com
Contact
Founder
Holger Schulze
Cybersecurity Insiders
holger.schulze@cybersecurity-insiders.com
