# Pwn2Own Berlin 2026: Hackers Discover 47 Zero-Days Worth $1.3M in Three-Day Exploitation Marathon


The Pwn2Own Berlin 2026 contest concluded this week with a staggering haul: security researchers earned $1,298,250 in cash rewards while exposing 47 previously unknown vulnerabilities across enterprise software, cloud platforms, and artificial intelligence systems. The three-day hacking competition, held at the OffensiveCon conference from May 14–16, once again demonstrated the abundance of exploitable flaws lurking in the world's most widely deployed software—and the immense value attackers place on discovering them.


## The Threat Landscape


The sheer volume of zero-day vulnerabilities discovered at a single event underscores a critical reality: enterprise software, despite continuous patching efforts, remains fundamentally vulnerable to determined attackers. The 47 zero-days found at Pwn2Own Berlin represent real-world attack vectors that malicious threat actors are actively seeking. Under the Zero Day Initiative's (ZDI) disclosure process, vendors now have 90 days to patch these flaws before public details are released—a window that dramatically increases the urgency for defenders to prepare.


What makes this event particularly alarming is the breadth of vulnerable systems: Microsoft products dominated the leaderboard, with Windows 11, Microsoft Exchange, and Microsoft Edge each compromised multiple times. But the vulnerability landscape extended far beyond Redmond—attackers also exploited Red Hat Enterprise Linux for Workstations, VMware ESXi, NVIDIA Container Toolkit, and multiple AI coding agents. The diversity of targets signals that zero-day vulnerabilities are not isolated edge cases; they are systemic across entire product categories.


## Background and Context


Pwn2Own Berlin has become the de facto championship for vulnerability researchers and security teams. The contest attracts the world's top defensive talent—teams like DEVCORE (Taiwan), STARLabs (Singapore), and IBM X-Force Offensive Research compete to demonstrate that they can compromise fully patched systems under controlled conditions. The cash rewards are substantial, but the prestige may matter more: a top finish at Pwn2Own is a credential that attracts venture capital, enterprise customers, and security researchers looking to prove their capabilities.


This year's event saw daily escalations in intensity and reward payouts:


| Day | Zero-Days Discovered | Cash Awarded |

|-----|----------------------|-------------|

| Day 1 (May 14) | 24 | $523,000 |

| Day 2 (May 15) | 15 | $385,750 |

| Day 3 (May 16) | 8 | $389,500 |

| Total | 47 | $1,298,250 |


DEVCORE captured first place with an astounding $505,000 in rewards across 50.5 Master of Pwn points, compromising critical infrastructure including Microsoft SharePoint, Microsoft Exchange, Microsoft Edge, and Windows 11. STARLabs took second with $242,500, followed by Out Of Bounds in third with $95,750.


## Technical Details: The Most Impactful Exploits


The contest's headline achievement came from Cheng-Da Tsai (Orange Tsai) of DEVCORE, who earned the contest's single largest payout: $200,000 for chaining three separate vulnerabilities to achieve remote code execution with SYSTEM privileges on Microsoft Exchange. This exploit is particularly dangerous because Exchange servers are typically internet-facing, meaning a remote attacker with no prior access could potentially compromise critical business communication infrastructure.


Tsai also earned $175,000 on the first day for demonstrating a Microsoft Edge sandbox escape by chaining four separate logic bugs. Sandbox escapes are particularly valuable to attackers because they allow exploitation of browser vulnerabilities to escape the normally restricted execution environment and gain full system access.


Other notable exploits included:


  • Multiple Windows 11 local privilege escalation (LPE) vulnerabilities demonstrated across multiple days, allowing unprivileged users to gain SYSTEM-level access
  • Red Hat Enterprise Linux for Workstations privilege escalation, shown multiple times throughout the contest
  • AI coding agent zero-days, highlighting emerging risks in machine learning-based development tools
  • VMware ESXi memory corruption exploit, enabling attackers to compromise virtualization infrastructure
  • NVIDIA Container Toolkit vulnerability, exposing container orchestration environments

    • IBM X-Force researcher Valentina Palmiotti (chompie) contributed $70,000 in rewards for rooting Red Hat Linux for Workstations and discovering an NVIDIA Container Toolkit zero-day, signaling that infrastructure-focused vulnerabilities command premium payouts.


    ## Implications for Enterprises


    The Pwn2Own results carry sobering implications for IT security teams worldwide:


    1. No software is truly secure in its current state. Even with security patches applied within days of release, fully updated systems can be compromised through chained vulnerabilities that bypass multiple layers of protection.


    2. Microsoft products remain the primary target. The dominance of Microsoft exploits at Pwn2Own reflects both the company's market share and the attention security researchers devote to breaking widely deployed software. Organizations running Windows, Exchange, and Office 365 should assume they are priority targets for sophisticated threat actors.


    3. Infrastructure is increasingly exposed. Exploits targeting VMware, NVIDIA, Red Hat Enterprise Linux, and container toolkits indicate that attackers are focusing on the underlying infrastructure that supports modern enterprises. A single compromised hypervisor or container orchestration platform could grant an attacker access to dozens or hundreds of virtual machines.


    4. AI systems introduce new attack surface. The presence of AI coding agent zero-days suggests that rapidly deployed machine learning systems are not receiving the same security scrutiny as legacy enterprise software, creating novel risks for organizations integrating AI tools into development pipelines.


    5. The 90-day patch window creates a critical vulnerability period. ZDI's standard 90-day disclosure timeline means organizations have roughly three months to patch these flaws after vendors receive notice. In practice, many organizations take significantly longer to deploy updates, leaving them exposed to determined threat actors.


    ## Recommendations for Defenders


    Immediate actions:


  • Inventory Microsoft products in your environment (Windows, Exchange, SharePoint, Edge) and prioritize patch deployment for any zero-days disclosed in the coming weeks
  • Review privilege escalation controls on Windows systems—many of the LPE exploits can be mitigated through AppLocker, Windows Defender Application Control, and restricted token policies
  • Audit hypervisor access and container orchestration platforms; ensure that management interfaces are not exposed to untrusted networks
  • Disable or isolate AI coding agents until vendors release security patches and you've conducted risk assessment
  • Enable enhanced logging on Exchange servers and other internet-facing systems to detect exploitation attempts

    • Longer-term strategy:


  • Adopt assume-breach principles: assume that sophisticated attackers will find and exploit zero-days despite your patching efforts
  • Implement network segmentation to contain the blast radius of any single compromise
  • Establish compensating controls (EDR, behavioral analytics, threat hunting) that detect and respond to exploitation attempts even when preventive patching fails
  • Consider zero-trust architecture to reduce the impact of privilege escalation vulnerabilities

    • ## HackWire Analysis


    Pwn2Own Berlin 2026 reveals a troubling truth that security vendors prefer to downplay: the patch gap is not closing, it's widening. While Microsoft and other vendors invest billions in security research, threat actors are outpacing them in discovering exploitable flaws. The $1.3M payout at a single conference represents not just lost opportunities for attackers to sell zero-days to criminal groups or state-sponsored actors—it represents a direct measure of how much low-hanging fruit still exists in "secure" enterprise software.


    The most concerning insight is the expansion beyond traditional endpoints. Previous Pwn2Own contests focused heavily on browsers and operating systems. This year's results show that infrastructure software—hypervisors, container toolkits, Linux distributions—has become equally valuable. This shift reflects the reality of modern enterprises: an attacker who compromises a hypervisor gains access to every virtual machine it hosts, potentially affecting thousands of users and systems. The container toolkit exploits are similarly multiplied: a single vulnerability in a widely deployed tool can be weaponized across container fleets spanning entire cloud infrastructure.


    The presence of AI coding agent zero-days also signals what may be the next major security headache for enterprises. These tools are being deployed at velocity that far outpaces security review. As with early web application firewalls and cloud platforms, the first generation of AI development tools are being adopted for productivity gains with security considerations coming as an afterthought. Pwn2Own researchers found exploitable flaws in these tools within months of their deployment—a pattern that will likely repeat as new AI systems proliferate.


    For defenders, the lesson is clear: assume that zero-days affecting your most critical systems will be discovered and exploited during your patch window. Build your detection and response posture around that assumption, not around the fantasy that patching alone will protect you. The researchers earning these rewards are skilled, but they're not superhuman. Criminal groups and state sponsors have equal or greater capabilities—they're just not submitting their findings to ZDI for public credit.


    — HackWire Editorial


    ## Related Coverage


  • Read more in our [Vulnerabilities](https://www.hackwire.news/category/vulnerabilities) coverage
  • Cross-reference with [Breaches](https://www.hackwire.news/category/breaches) and [Malware](https://www.hackwire.news/category/malware)
  • Stay current via the [HackWire homepage](https://www.hackwire.news/)