With AI adoption spiking, enterprises need to begin considering data in the AI world. This means not only what data is being generated, but also questions of management – how it is categorized, how you can determine and track access, and how we can maintain productivity and security.
There’s an inherent contradiction in pressures being experienced. On the one hand, agentic AI has boosted productivity by massive levels in various industries. Where once AI had primary value in analysis and surfacing insights for humans to interact on, now agentic AI is beginning to take on an active role in execution.
This, naturally, requires access to internal data and the ability to manipulate it, which runs directly into questions of safety and accountability. IBM’s 2025 data claims data breach costs of 4.4M USD on average, taking into account direct and indirect costs and damages.
Right now, the same data shows that many organizations are sidelining data management in the name of cashing in on those productivity boosts, accepting the increased degree of risk.
What are the actual concerns?
The main issue being discussed is effectively one of credentials; security, including measures such as zero-trust, have leaned on being able to identify and validate identities online including their right to access and manipulate given data. Likewise, audits use this as a jumping off point to identify points of concern.
Currently operational controls for non-human identities are still not as strongly developed. Implementing agentic AI workflows has, in some cases, led to companies discovering previously unconsidered issues including API misuse or credential abuse.
The solution, into the future, is going to have to include not only existing security measures like Zero Trust and immutable archive and backup storage but also a comprehensive way to identify automated systems, classify them and record their actions within your system to allow for future audit and regulatory compliance. Developing this solution is going to be the new data security frontier that defines the field into the latter half of this decade.
Your Data in Your hands – With TECH-ARROW
by Matúš Koronthály
Image generated by Canva
