KorvaJoin the waitlist
Tech

Test What Attackers Can Reach Before You Scan All Your Code

A Wiz chief technology officer says unauthenticated internet exposure matters more than code flaws, and should be tested continuously.

AI image
The a16z Show · Every Day Is Zero Day | Wiz CTO Ami Luttwak

Ami Luttwak, chief technology officer of Wiz, has a blunt rule for companies worried about AI attackers: attack yourself first, and do not stop.

His company's red agent does exactly that, probing a customer's own applications from the outside while using inside context to find weaknesses before adversaries do.

The logic is that defenses cannot stay fixed when attacker skills improve every week. He compares it to hiding a door key under a mat and calling it secure, until a stronger intruder learns to look there.

Instead of spending a whole budget scanning code, he recommends a staged pipeline. First map exposure in a straightforward, deterministic way, then check everything that is open to the internet without a login.

That unauthenticated surface comes first because that is where attackers usually enter. He points to a recent incident in which a hard coded key in a mobile app exposed a large amount of people's data.

Only then spend more on deeper work. A routine endpoint might get a few minutes of automated attention, while a critical application can merit ten hours of attempted hacking.

When engineers say a fix is too costly, the finding is not dropped. It is handed to a defensive agent to add monitoring around that spot.

The first step for a busy team, then, is to list what is reachable without authentication and test it continuously, before paying to comb every line of code.

Questions people ask

What counts as unauthenticated exposure?

It means anything an outsider can reach without signing in, such as a website, an application programming interface, a data bucket, or a login page. Attackers prefer these paths because they require no stolen password and can often be tested at scale.

How often is continuously in practice?

The source describes it as ongoing testing rather than a one-time check, so new weaknesses are found soon after they appear. The point is to keep testing your own systems with the latest tools instead of relying on a single review.

What does the defensive agent do?

It is the defensive counterpart to an attack simulator. It watches for suspicious activity, investigates alerts, and can add monitoring or help contain a problem after the attacking tool finds a weakness that is hard to fix quickly.

More in Tech