Vulnerabilities

Hackers Win $260,000 on First Day of Pwn2Own Berlin 2025

Pwn2Own participants have earned tens of thousands of dollars for Red Hat, Windows, Oracle VirtualBox, Docker Desktop, and AI exploits.

Pwn2Own Berlin 2025

The Pwn2Own Berlin 2025 hacking competition organized by Trend Micro’s Zero Day Initiative (ZDI) is taking place these days in Berlin, Germany, and on the first day participants earned a total of $260,000.

Pwn2Own Berlin 2025 introduces AI hacking and the first ever winner in this category is Sina Kheirkhah of Summoning Team, who earned $20,000 for an exploit targeting the Chroma open source AI application database.

The same researcher earned an additional $15,000 for a different exploit in the same category. He successfully hacked an NVIDIA Triton Inference Server, but it has been marked as a ‘collision’ because the vendor had known about the bug but had yet to patch it.

Viettel Cyber Security earned the same amount for another known NVIDIA Triton vulnerability that had yet to be patched by the tech giant.

The highest single reward was earned on the first day by the Star Labs team, which received $60,000 for an exploit chain that involved a Linux kernel vulnerability to perform a Docker Desktop escape and ultimately execute code on the underlying operating system. 

Team Prison Break got $40,000 for escaping Oracle VirtualBox and executing code on the underlying OS. 

Advertisement. Scroll to continue reading.

Others earned between $15,000 and $30,000 for Red Hat and Windows 11 exploits.

Over the next two days, Pwn2Own Berlin 2025 participants will attempt to demonstrate hacks against NVIDIA Triton Inference Server and Container Toolkit, Microsoft SharePoint, VMware ESXi and Workstation, Firefox, Redis, VirtualBox, Red Hat, and Windows 11.

Related: Exploited VMware ESXi Flaws Put Many at Risk of Ransomware, Other Attacks

Related: Hackers Earn $886,000 at Pwn2Own Automotive 2025 for Charger, OS, Infotainment Exploits

Related: $2.5 Million Offered at Upcoming ‘Matrix Cup’ Chinese Hacking Contest

Related Content

Artificial Intelligence

The startup maps how identities, assets and vulnerabilities connect, and uses AI agents to find and fix risky combinations of issues.

Artificial Intelligence

Google’s new frontier model scored 88.2% on SEC-bench Pro and is rolling out first to vetted cyber defenders via its Fairwind Program.

Government

Unpatched voting machines, unverifiable QR-coded ballots, and AI-driven influence campaigns leave US elections short of secure.

Cybersecurity Funding

The company will use the Series B funds to expand its platform for securing AI agents’ access to sensitive systems.

Artificial Intelligence

Anthropic presents its own initial public offering as a step toward mitigating the risks it believes AI poses.

Artificial Intelligence

Xint says a flaw in OpenAI’s compaction process let researchers move reasoning traces between models, potentially making it cheaper to distill a rival AI.

Artificial Intelligence

The model engaged in severe overreach and dishonesty, creating fake identities and submitting malicious code on GitHub.

Artificial Intelligence

The state-sponsored hackers targeted hundreds of organizations and gained privileged access to some environments.

Copyright © 2026 SecurityWeek ®, a Wired Business Media Publication. All Rights Reserved.

Exit mobile version