- At Pwn2Own Ireland 2026, researchers exploited 98 zero-days and earned $1,262,000, according to BleepingComputer's tally of the contest run by Trend Micro's Zero Day Initiative.
- Contest rules require the latest firmware, so being fully patched did not stop these exploits. Patch level shows what is fixed, not what is already found.
- Ask vendors how fast they fix flaws and whether independent researchers have tested their AI and device products.
A fully updated device is not the same as a finished one. Patching removes the flaws that are known. It says nothing about the ones a skilled researcher is about to find. Pwn2Own Ireland 2026 made that gap easy to see.
Up to date, and still open
Trend Micro's Zero Day Initiative (ZDI) runs Pwn2Own. According to BleepingComputer's report on the contest, the rules require every target to run the latest firmware. Contestants must also show arbitrary code execution, meaning they can run their own commands on the device.
Over three days, 29 research teams took on products in seven categories. BleepingComputer reports they exploited 98 zero-days and collected $1,262,000. A zero-day is a flaw the vendor has had no time to fix.
The final day alone brought 21 zero-days and $641,000. The top prize was $300,000, won by Ikotas Labs for chaining several bugs to take over a Google Pixel 10. Ikotas Labs finished first overall with $361,000 and 42.5 Master of Pwn points.
ZDI's results label that Pixel 10 win a success with a collision. The post does not say how many of the chained bugs were new. So even the top prize does not map cleanly onto a count of zero-days.
Chains matter, but count the zero-days carefully
Many winning entries were chains. A chain links several small flaws, each too minor to matter alone, into one path to full control. A skimmer should not read every "five-bug chain" as five unknown flaws, though.
ZDI's day-three results show the difference. Team DDOS took over a Home Assistant Green with a five-bug chain that included just one zero-day. Team MAMMOTH later beat the same device with a chain of six zero-days. Summoning Team's five-bug attack on a Philips Hue Bridge Pro was labelled a full collision, not a zero-day chain. ZDI's post does not define "collision". Read in context, it marks bugs that were not new to the contest.
For a manager, the lesson is still that a score on a single bug can understate the risk. The real exposure is what an attacker can build from several low-rated flaws. Some of those flaws may already be known, and some may not.
The devices around your people
Day three also covered Brother, Canon and Lexmark printers. Each was exploited. Printers sit on office networks, and smart-home hubs sit in homes and sometimes in workplaces. Nobody thinks of them as high-value systems, yet they ran the latest firmware and still fell.
Apple's iPhone 17 was a possible target with a $300,000 maximum award for a remote hack. BleepingComputer reports that no contestant registered an attempt. The reports give no reason, so it tells us nothing about the phone's security.
AI products are now on the target list
The 2026 categories included AI infrastructure and AI coding apps. Ikotas Labs, the overall winner, hit two AI products on its way up the table, per BleepingComputer: OpenAI Codex and Oracle's Autonomous AI Database. Two more teams, OtterSec and Platform Security, also exploited the Oracle database on day three, with chains of four and five bugs.
These are contest payouts, set by the organiser. They are not a measure of how risky each product is. They do show that AI databases and coding tools now get the same scrutiny as phones. The ZDI post also lists a scheduled attempt on Chroma. The sources do not report its result.
What collisions do to the count
ZDI labelled most day-three successes as collisions. OtterSec's Oracle chain had three collisions and one zero-day. Platform Security's chain had four collisions and one unique bug. The 98 headline figure is BleepingComputer's count of zero-days, and the sources do not show how it treats these mixed chains.
BleepingComputer also reported that on day one, Samsung already knew some of the bugs exploited in the Galaxy S26. Together these suggest some overlap between what teams found and what vendors had in hand. The sources do not say how much.
The 90-day clock and what to ask
BleepingComputer reports that vendors must patch zero-days disclosed at Pwn2Own within 90 days before ZDI shares details publicly. That is a fixed deadline, and organisations can ask their own vendors whether they work to anything similar.
Leaders can put four questions to their teams. First, which printers, hubs and phones on our networks are on this year's target list? Second, how quickly do those vendors ship fixes, and how do we learn of them? Third, have independent researchers tested the AI databases and coding tools we buy? Fourth, do we plan around chains of small flaws, or only around the single high scores?
Staying up to date is necessary. It is not proof that nobody has found a way in.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Trend Micro's Zero Day Initiative (ZDI).
