On-Demand Videos & Webinars

Shadow IT with autonomy: Donna Ross on what unsanctioned AI really changes

Donna Ross started in data center operations at Prudential, back when it wasn't called cyber. What the jobs after it had in common, she says, is building programs that didn't exist before. She stood up the security program for GMAC's first bank, then went to Corning, then a year at a startup, then ten-plus years as CISO of a financial services company. She's a month into retirement and busier now than the job ever made her.

Mitiga Field CISO Brian Contos asked her whether unsanctioned AI is simply shadow IT with a larger blast radius. A little of each, she told him. Decades of unsanctioned SaaS taught her the arrival pattern, and it hasn't changed: the tools get there before security knows about them or has controls around them. The difference, she says, is that AI changes the risk profile entirely.

"Shadow IT used to expose data. Unsanctioned AI exposes decisions."

A model doesn't store information only, she says. It acts on it, writing code and triggering workflows across the enterprise at a speed and scale no person works at. A junior analyst can deploy an agent that behaves like a senior engineer because the agent has the same access as the analyst. A marketing team can spin up an agent that touches customer data without putting in a single ticket.

"This is bigger than shadow IT. It's shadow IT with autonomy."

From a board seat, she sees both what gets asked and what should be asked. What boards tend to ask is whether the company is using AI safely, which she considers too broad to act on. Her question is permissions. Where in the environment can an autonomous agent take action with no human in the loop, what can it do at two in the morning when nobody's watching, and are there guardrails?

"Stop asking if we are using it safely and start asking about permissions."

She is just as direct about how CISOs lose the room. Don't say model drift to a board. Say what model drift does to loan decisions. Don't say SaaS sprawl. Say that SaaS sprawl increases regulatory exposure into customer workflows. Boards don't need metrics or mechanics, she says. They need consequences.

Which is why she thinks the job is changing. The next generation of CISOs gets hired for influence, not packet capture, and may well come from risk, audit or legal. Her shorthand for the shift is a pun she has clearly used before: getting from the Doctor No to the Doctor Can-Know.

The weight of the seat she describes is plain. She didn't realize how much pressure she was carrying until the pressure valve got released. You're accountable for systems you don't own, decisions you don't make, and budgets you don't approve, and if you work for a public company, you can be personally liable for outcomes you can't fully control. Her answer as a leader was work-life harmony, not balance. If someone on the team asked for two hours to go to the doctor, she'd ask them to take the day.

When asked what she'd hand the next generation, she keeps it short. You can't secure what the business doesn't understand. Your first job is education, your second is alignment, and the technical work comes third.

Don't miss these stories