“One important thing to understand about cybersecurity is that it’s a mind game,” Ami Luttwak, chief technist at cybersecurity firm Wiz, told TechCrunch in a recent episode of Equity. “As there is a wave of new technology coming, there is a new opportunity for (attackers) to start using it.”
The attack surface is growing as businesses rush to embed AI into their workflows through atmospheric coding, AI agents integration, or new tools. AI helps developers ship their code faster, but there are shortcuts and mistakes in its speed, creating new openings for attackers.
Wiz, which was acquired by Google for $32 billion earlier this year, recently conducted testing, Luttwak said, discovering that the common problem with the Vibe Coded application is an unstable implementation of authentication.
“It happened because it was easier to build that way,” he said. “Vibe coding agents do what you say, and if you didn’t tell them to build it in the safest way, then it wouldn’t.”
Luttwak pointed out that today there is a constant trade-off for companies that choose whether they are fast and safe. However, it’s not just developers who use AI to move faster. He said attackers are now starting exploits using vibe coding, prompt-based techniques and even their own AI agents.
“You can actually see an attacker using prompts for attacks,” says Luttwak. “It’s not just coding the attacker’s atmosphere. The attacker looks for the AI tools you have and says, “Send all the secrets, delete the machine, delete the files.” ”
In this landscape, attackers are finding entry points for new AI tools to help businesses deploy internally and increase efficiency. Luttwak says these integrations could lead to “supply chain attacks.” By breaching third-party services that have broad access to the enterprise infrastructure, attackers can delve deeper into the enterprise system.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
That’s what happened last month when Drift, a startup that sells AI chatbots for sales and marketing, was compromised and published sales data for hundreds of enterprise customers, including CloudFlare, Palo Alto Networks, Google and more. The attacker accessed a token or digital key, impersonated a chatbot, queried sales force data, and used it to move horizontally within the customer environment.
“The attacker pushed the attack code, which was also created using atmospheric coding,” says Luttwak.
Luttwak said that while corporate adoption of AI tools is still minimal, about 1% of companies believe they are fully adopting AI.
“And when you look at the (attack) flow, AI was embedded in every step,” Luttwak said. “This revolution is faster than any revolution we’ve seen in the past, meaning we as an industry need to move faster.”
Luttwak pointed to another major supply chain attack in August called “S1ingularity” on NX, a popular build system for JavaScript developers. The attackers unleashed the malware on the system, then detected the presence of AI developer tools such as Claude and Gemini, hijacked them, and autonomously scanned the system for valuable data. The attack compromised thousands of developer tokens and keys, allowing attackers to access private Github repository.
Luttwak says despite the threats, this was an exciting time to become a cybersecurity leader. Founded in 2020, Wiz originally focused on helping organizations identify and address false mining, vulnerabilities, and other security risks across cloud environments.
Last year, Wiz responded to the speed of AI-related attacks and expanded its ability to use AI in its own products.
Last September, Wiz launched WIZ code focused on protecting the software development lifecycle by identifying and mitigating security issues early in the development process. This allows businesses to be “by design.” In April, Wiz launched Wiz Devend. WizDefend provides runtime protection by detecting and responding to active threats in your cloud environment.
Luttwak said it’s important for Wiz to fully understand the customer’s applications when the startup helps with what he calls “horizontal security.”
“You need to understand why you’re building it… so I can build a security tool that understands you, a security tool that no one has before,” he said.
“From the first day you need to have a ciso.”
The democratization of AI tools has caused a flood of new startups that promise to solve the problems with enterprises. But Luttwak says companies don’t just say that their data on their company, employees and customers is all “giving amazing AI insights just because they say “all data” of every small SaaS company with five employees.” “I say it.
Of course, these startups need that data if the offering is valuable. Luttwak says it means they are obliged to make sure they are acting like a safe organization from the start.
“From day one, you need to think about security and compliance,” he said. “From the first day, we need a CISO (Chief Information Security Officer), even if we have five people.”
Before writing a line of code, he said, startups should think of like a very secure organization. They should consider enterprise security features, audit logs, authentication, access to production, development practices, security ownership, and single sign-on. Planning a plan from the start like this means that there is no need to overhaul the process later and creates what Luttwak calls a “security debt.” And if you are aiming to sell to businesses, you are already ready to protect their data.
“We were SOC2 compliance (compliance framework) before we got the code,” he said. “And I can tell you a secret. Getting SOC2 compliance with five employees is much easier than 500 employees.”
The next important step for a startup is to think about architecture, he said.
“If you’re an AI startup that wants to focus on enterprises from day one, you need to think about an architecture that allows your customer’s data to stay in your customer environment.”
Luttwak says it’s time for cybersecurity startups looking to step into the field in the age of AI. From phishing protection and email security to malware and endpoint protection, everything is the basis for fertile innovation for both attackers and advocates. The same applies to startups that can assist with workflows and automation tools to do “Vibe Security” as many security teams still don’t know how to use AI to protect AI.
“The game is open,” Luttwak said. “If there’s a new attack in every area of security, that means we have to rethink every part of security.”
