Google’s $32B Bet on Cybersecurity in the AI vs. AI Era | Wiz Co-Founder Yinon Costica

Summary · 3 min read

Costica previously helped build another cybersecurity company and later ran Microsoft's cloud-security work. He co-founded Wiz, which Google agreed to acquire for $32 billion, described as the largest acquisition in Google's history.

Costica says Wiz emerged from customer interviews that revealed a cloud-security gap driven as much by process and people as by technology. He credits Wiz's rapid growth to an enterprise-focused product that helped security teams prioritize problems and developers fix them. He argues that AI gives attackers an immediate advantage, but can help defenders test and secure more of their environments if organizations adopt it quickly. He says trust, agent guardrails and overcoming organizational silos remain central, while humans should continue to set security goals.

Minute by minute

1. 2:11

   Real-world problems built the founding team. Costica says he and his co-founders met at 18 in Israel's Talpiot program, where they studied math, physics and computer science before entering cyber work. He credits real-world technical tasks and teamwork with preparing them to build companies, and says trust and mutual respect helped sustain their partnership.

2. 6:28

   Customer interviews redirected Wiz. After five years at Microsoft, Costica and his co-founders left in 2020 intending to build cloud networking. Customer interviews during COVID showed that buyers struggled to identify deployment, ownership and budget for networking, but repeatedly raised cloud visibility and security problems.

3. 11:27

   Cloud security needed a new process. Costica says cloud adoption during COVID created a market window, while Wiz's enterprise-ready product supported rapid growth. He says the industry had focused on available technology but missed the people and process problem; Wiz helped security teams prioritize issues and developers fix them.

4. 14:09

   Scale requires constant alignment. Costica says fast-growing companies need a clear shared way to explain the problem, solution and value, updated as the business changes. He credits Wiz's flat structure and regular customer feedback with helping the company adapt while keeping teams aligned.

5. 17:14

   Atlas automates vulnerability research. Costica says Atlas breaks vulnerability research into tasks such as analyzing code, finding entry points, exploiting vulnerabilities and validating results, using different models for different tasks. He says it found hundreds of vulnerabilities in heavily audited open-source repositories and can work across many repositories in minutes or hours rather than months or years.

6. 22:42

   Defenders must adopt AI faster. Costica says AI can make it easier for an attacker to carry out one attack than for defenders to secure an entire estate with new tools. He argues that AI could help defenders scan code continuously, but says they need to accelerate adoption before attackers gain the advantage.

7. 27:28

   Consumer agents need guardrails. Costica says agents can make technology easier to use, but people should consider permissions, audit logs, data handling and who provides the service before granting access to sensitive accounts. He says an ecosystem of rules and responsibilities will need to develop, and that lasting trust will matter.

8. 34:41

   Models add supply-chain risks. Costica says open-weight models may be useful for specific tasks, but users need to scrutinize their sources and test how they behave. He says Wiz found a DeepSeek vulnerability in the underlying infrastructure that exposed large data sets, not in the model itself.

9. 38:32

   AI can strengthen defensive readiness. Costica says companies should use AI to test their own environments before attackers do, since defenders have more internal context. He also says AI can speed up identifying attack patterns and sharing indicators across environments, while task-specific defensive models could add another advantage.

10. 43:21

    Google expands Wiz's reach. Costica says Google brings Wiz threat intelligence, access to models and computing capacity, and broader distribution. He says Wiz began experimenting with assistive AI early and, after reframing its approach, built red, blue and green agents in three to four months.

11. 47:13

    Security adoption remains the bottleneck. Costica says the biggest unsolved problem is implementing AI security across organizations, including building trust in agentic workflows and reducing approval delays. He says teams responsible for applications, cloud, networks and identity need a more unified view and process.

12. 49:50

    Agent security needs guardrails. Costica says an agent can expose sensitive data even when it follows instructions, so organizations should limit access, audit activity and inspect prompts and responses. He says Wiz's red agent finds issues at almost every customer that enables it, including paths human researchers had not found.

13. 53:53

    Software must serve agents. Costica expects more software to be consumed through agents rather than user interfaces, with SaaS products making their data models available to both. He says websites may need agent-readable access, while humans should remain the moral compass that sets goals and judges security outcomes.

14. 58:21

    AI could make security more scalable. Costica says it is too early to identify AI's winners because lasting customer value and trust take time to build. He argues that AI could remove security bottlenecks and make defense more scalable, but says the speed of adoption matters in the race against emerging threats.

Published Oct 7, 2026.