Most teams accept higher risk for faster AI database work
Database professionals are using AI for everyday work like writing queries, building schemas, and reviewing code, and a growing share rely on autonomous tools that act on the database itself. The use of AI in database management has almost tripled in a year, climbing from 15% to 44% of organizations, according to Redgate’s 2026 State of the Database Landscape report. That puts AI inside the systems holding an organization’s most sensitive data, often with permission to change that data directly.
The security problem here is one the people running these systems already see coming. Data security ranks as their top worry about AI, named by close to two-thirds of respondents and by even more of those whose organizations have yet to adopt it. The same group is going ahead anyway. A majority accept higher data security risk in exchange for the speed AI gives them, a choice many made before the controls were in place to support it.
Autonomous tools with standing access
Most of the attention on AI in this space goes to generative tools that draft code and queries. The sharper issue is the next category. A majority of organizations use autonomous AI and agents, software that can operate on database systems with limited human review of each action. These tools handle data quality work, schema design, and automation, and almost every organization reports getting something useful out of them.
“AI can be the good and the bad here. It can analyze threats and find issues automatically, so it can strengthen my own system’s security. The concern is mainly how people deal with it (like uploading confidential information to ChatGPT), or when you give it too much power. Granting AI the permissions to automatically fix everything means it can potentially also cause huge damage,” said Ben Weissman, CEO, Solisyon.
Weak controls underneath fast change
The foundations that would contain that risk are thin. Most organizations still depend on manual steps to test and deploy database changes, few use formal data governance frameworks, and metadata management remains rare. Monitoring is spread across a mix of vendor products, scripts, and home-grown solutions. AI speeds up how fast database code and schemas change, and faster change makes mistakes easier to introduce and harder to catch before they hit production.
Data quality compounds the problem. A large share of respondents report quality issues surfacing when data moves between systems, the exact conditions where automated changes do the most harm. Gartner has tied many AI failures to unready data and predicts that most projects lacking AI-ready data will be abandoned by the end of 2026.
Shadow AI in the data layer
Much of this AI use sits outside any defined workflow. Individuals pick their own tools to move faster, and those habits stay informal and ungoverned. The pattern looks like earlier waves of shadow IT, now operating against the data store. Weissman pointed to the everyday version of it: people pasting confidential information into tools like ChatGPT. A minority of respondents remain uneasy about feeding company data into AI at all, and a small share of organizations have banned it outright. Many more limit use to an approved list, and that approved-only group is growing.
Some organizations are adding structure. Most now provide formal guidance on AI use, a sharp rise from a year earlier. The technology is changing teams too. Around half of respondents say their actual work has drifted from their job description, and a similar share of organizations are hiring fewer junior staff because of AI.
Exposure set to grow
The number of AI tools in use will keep climbing. More than eight in ten respondents expect to bring on additional AI tools in their roles over the next one to two years, with growth concentrated in large, cloud-based environments that hold the most data. Spending backs that up, with heavy investment from large enterprises in particular.
The questions security practitioners need to ask are old ones about identity, least privilege, auditability, and rollback, now pointed at a new actor with elevated rights over sensitive data.
“AI is accelerating at a pace the world has never seen, but it’s only as powerful as the data that feeds it and databases are the crown jewels of the intelligence space. The data professionals who guard and govern these data ecosystems will determine whether AI becomes humanity’s greatest innovation or its most consequential mistake,” said Kellyn Gorman, Multi-platform and AI Advocate at Redgate Software.
Demo: Prophet Agentic AI SOC Platform transforms alert triage and investigation