1st Place at Palo Alto Networks' AI Hackathon: Meet Team SafeContext

Blog | Innovation & Impact | Career Development | Culture

Friday, April 17, 2026

Embracing Innovation at Our Largest Internal Hackathon

In February 2026, Palo Alto Networks hosted its largest-ever internal AI hackathon — the Agentic AI Hackathon — in partnership with Google Cloud and Gemini Enterprise. Over two days, 70 teams and more than 200 engineers from across Palo Alto Networks came together to explore how agentic AI can be applied to solve real-world cybersecurity challenges.

Teams competed across three preliminary tracks, with nine finalists advancing to the final round. Google engineers served as judges alongside Palo Alto Networks leaders. Projects spanned a wide range of cybersecurity applications — from IoT security and malware analysis to software supply chain protection and AI-powered DevOps. The breadth and quality of submissions led the hackathon organizer to note that the ideas presented would "impact our product roadmap for the next three to five years."

At a Glance: Featured Profile

Featured Profile: Alexander, Principal Machine Learning Engineer; Lauren, Staff Machine Learning Engineer; Qinge, Senior Staff Researcher; and Nabeel, Senior Principal Researcher
The Mission: AI Agent protection
The Result: SafeContext, a program protecting AI agents from web-based threats
The Takeaway: As we face emerging threats, innovation is more essential than ever

Keeping AI Agents Safe with SafeContext

Our team, "Stop Trying to Make Fetch Happen," took 1st Place in the Idea Exploring track. With the support of our team advisor Hongliang, we used our combined expertise in URL reputation, threat intelligence, and content security to tackle AI agent protection. The result was SafeContext, a project focused on protecting AI agents from emerging web-based threats like indirect prompt injection. The SafeContext prototype was built using Google's Agent Development Kit (ADK), putting the Palo Alto Networks and Google Cloud partnership into action.

Why AI Agent Security Matters

As AI agents become more deeply integrated into browsers, developer tools, and enterprise workflows, they introduce a new class of security challenges. Unlike humans — who read web content as text — AI agents can interpret that same content as executable instructions.

This isn't a theoretical concern. As documented by Palo Alto Networks' Unit 42 research team in their recent publication "Fooling AI Agents: Web-Based Indirect Prompt Injection Observed in the Wild," indirect prompt injection attacks are being actively weaponized in the wild, with 22 distinct attacker techniques already identified. Our team member Nabeel co-authored that research — and the SafeContext project was born from that same deep understanding of the threat landscape.

Innovation Culture in Action

Hackathons like this are a reflection of how Palo Alto Networks invests in its engineering talent — creating space for teams to explore bold ideas, collaborate across organizations, and prototype solutions to emerging security challenges. Competing against 70 teams and taking first place was an incredible experience, and we're proud to be part of a company that values this kind of innovation.

Quotes:

"Competing against 70 teams and earning 1st place was an incredible experience. It shows how much innovation happens when you bring together people with different expertise and give them space to experiment." — Alexander, Principal Machine Learning Engineer

"It is the context that makes an agent great. And that also means it is malicious context that makes an agent malicious. A malicious command to us people is text on a screen. But to an agent, it is an executable — one that could wipe your entire hard drive before you've finished your morning coffee." — Lauren, Staff Machine Learning Engineer

"It was an unforgettable experience working with the team during the AI hackathon. Exploring how AI agents can be practically applied in cybersecurity is exciting, and it gives us a real look at next-gen security in action." — Qinge, Senior Staff Researcher

“Innovation driven culture here at Palo is next level. This agentic AI hackathon was a perfect example - A chance to work with a great team on a problem that actually matters: how can enterprises leverage AI agents securely? Turns out, the answer is to use more AI!” – Nabeel, Senior Principal Researcher