AI is becoming a working tool in security operations centres, but many teams remain uneasy about giving it room to act. A Dropzone AI survey cited by ITPro found 39% of respondents used AI in full production, while the same share reported using it for autonomous investigations.
Watch Desk analysis
What happened
The survey found that 55% of security alerts were never investigated or validated, while just 2% of organisations said they investigated every alert. Dropzone AI international vice-president Brett Candon told ITPro that AI could help address this workload, but warned that simply adding AI is not enough.
Trust remains a practical barrier: 25% of respondents cited data security and privacy concerns, and the same share worried that incorrect AI investigations or actions could affect security outcomes. Candon advocated a controlled approach. He described Dropzone AI’s agents as operating in read-only, investigate-only mode, collecting evidence without privileges to overwrite information or take action.
Why it matters
The appeal is plain: security teams face more alerts than they can reliably examine, and automation could help them investigate at machine speed. But an agent that can inspect sensitive systems is not just another chatbot. What it can access, and whether it can change anything, are central questions for the people responsible for keeping those systems safe.
The numbers come from a survey by Dropzone AI, a company selling security AI, and should be read in that context. Still, the figures point to a useful distinction: broad AI adoption does not mean teams are ready to let agents act without close limits.
Our read
The sensible test is not whether a security team uses AI, but whether it can show what the system is allowed to touch and what happens when it gets something wrong. Read-only investigation is a more measured first step than handing an agent the keys and hoping it has a quiet evening.
What to watch
- Whether security teams move from AI-assisted investigation to agents that can take actions.
- How providers define access limits, human approval and audit trails.
- Whether future surveys show growing trust alongside wider adoption.
Discussion spark: Should security teams let AI agents investigate alerts autonomously if they are read-only, or should a person approve every investigation?
Sources and evidence
- Source update (11 November 2026, 10:52 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.