It used to take a week or two for attackers to scan the internet and exploit a newly published vulnerability. Now it can happen within a day.
That changes what leadership teams need to pay attention to, because the attacks aren’t new. What’s new is the pace at which these attacks are happening.
Most of what we see is familiar, and AI has helped attackers move faster.
Like everything else, there are two sides to the coin. While AI helps your own teams ship code faster and helps your employees get work done faster, it also often does so with tools nobody approved.
The instinct is to answer speed with more tools. But tools don't decide what your people are allowed to do, which fix gets done first, or who owns the risk when something slips through.
People do—especially leaders.
We'll examine what that looks like in practice including what the latest data says, rules that hold up, how to govern AI without the theater, and how to make an honest call on what to build and what to buy.
The 2026 Verizon Data Breach Investigations Report looked at more than 22,000 confirmed breaches. Four findings stand out for leaders.
The human element is still present in 62% of breaches.
Here's the part that matters most. Even the AI-assisted malware Verizon observed mostly relied on well-known techniques. Fewer than 2.5% of cases used anything less common.
Metaphorically speaking, most companies have a lot of stuff in the basement. AI is turning on all the lights and going through it.
Again, the problems aren't new.
We're just seeing them more clearly and much faster. That's what the remediation numbers are telling us. And the most pressing matter at hand is getting it fixed before someone else finds it.
When everything speeds up, it's tempting to reach for a shortcut. We've written about most of them before, so here's the short version.
Each of these moves responsibility somewhere out of sight. But for your security measures to hold up, you need to keep responsibility in view.
Think of the ice cream in the fridge. Maybe you're not supposed to take it, but you do anyway. It's right there, it tastes good, and the consequences of “just this once” are small.
AI tools work the same way at work.
They're tempting, they're everywhere, and it takes seconds to paste something into one. The old-school "this is forbidden" security doesn't hold up against that. It's also why outright banning tools or AI doesn’t work, which brings us to our point about shadow AI before.
So the test for any security rule is simple.
Can people actually follow it in the middle of a busy workday?
If not, it's not a rule. It's a wish.
Govern AI without the theater
The public conversation about AI is stuck in a pro-versus-against debate. It's not a useful one. AI is here, your people are using it, and the real question is how.
At Aftra, we chose to step onto the AI train carefully with our safety boots on. Here's what that looks like for us, and what we'd suggest to any leadership team.
Weigh value against risk before you adopt
Before bringing in an AI tool, sit down and do the exercise. What are you hoping to get out of it? What's realistic? What are you putting at risk?
It sounds obvious. But it's easy to skip when everyone around you seems to be moving fast.
Let AI suggest. Let people decide.
In our own product, AI suggests and pre-fills. But a person always makes the final call.
Internally, we draw hard lines where the stakes are high. We don't allow AI agents to deploy code.
And because code is now being written faster than ever, we push to keep our codebase something humans can oversee by reusing components rather than generating new ones for every task.
Think of a Formula 1 car. It's the most capable car on the track. In the hands of an untrained driver, you won't even get it started. AI is the same. Much of its value depends on the person directing it.
A lot of what passes for AI governance is performance:
Guardrails only matter if they're operationally actionable.
If a rule can't be followed or checked, it's theater.
More and more vendors are adding AI features. Before you switch one on, ask:
And bring a second pair of eyes into the decision. (For reference, here's how we approached data privacy for Askja, our own AI assistant.)
None of this requires leaders to become security experts.
In our experience, ego is rarely what holds them back. Most are quick to say security isn't their specialty, and that's fine.
What helps is being brave enough to ask what feels like a stupid question. Often, it's the first time anyone in the room has asked it.
We've covered the practical side in depth elsewhere:
And one thing that's easy to forget. Use metrics to give credit, not just to find problems.
IT teams rarely get thanked because everyone expects the lights to be on. If you're measuring the right things, you can celebrate progress the way sales celebrates a new deal.
Imagine an AI that could read all your email, understand your contracts, and know your usual payments well enough to spot every business email compromise attempt. Essentially, it could make you immune to phishing.
The cost? Everything you send and receive gets scrutinized.
Would that trade be worth it?
There's no single right answer. It depends on your risk appetite. But questions like this are already on the table, and they aren't IT questions.
They’re questions for leadership to answer.
The attacks may not be new, but the pace is. You can’t win this one with more tools. But you can keep up, or at least try your very best to, by deciding clearly what the rules are and making sure people are able to follow them.