Follow Us on Google
By Christopher Oji
A United States-based 23-year-old tech developer, Tony Kabilan Okeke, has led a five-man team of Drexel University alumni and students in developing Xploit, an automated cybersecurity testing tool for AI agents designed to address growing security concerns in the global AI landscape.
Besides Tony Okeke, who is the team lead, other members are Kamdi Okeke, Kiitan Fawole, Dalu Okonkwo and Michael Moemeke.
Speaking with our reporter on the development, Okeke said: “As more businesses deploy AI agents that can take actions and use tools on behalf of customers, these systems become potential security risks. Unlike simple AI assistants, agents have access to tools and can perform real actions, meaning a security vulnerability isn’t just a PR problem; it could have serious real-world consequences.”
The team envisioned a tool that could automatically test AI agents for vulnerabilities by essentially playing the role of a digital attacker to identify weaknesses before real threats exploit them.
This idea emerged during a brainstorming session on November 21, 2025, when Tony led the group to build and pitch Xploit at the “Start-Up In a Weekend” Hackathon hosted from November 21 to 23, 2025, in Philadelphia by The Foundry & Velric, a Philadelphia-based founder-first community that serves as a startup ecosystem catalyst.
Okeke designed the system’s architecture and created the initial prototype of the user interface (UI). The UI concept was crucial because it needed to visually show how the automated attacker was thinking, strategising and attempting different approaches in real time through interactive graphs displaying the attack process as it unfolded.
Responsibilities were strategically divided among the team members. Some created sample AI agents to serve as “victims” for testing, Tony developed the core attacking system, one member refined the user interface, while others handled the technical infrastructure connecting all the components together.
The attacking system works like a strategic game player. It first selects an attack strategy, creates a detailed plan, executes that plan step-by-step by sending messages to the target AI agent, and analyses the responses to determine whether to continue or adopt a different approach.
Throughout the process, the web interface displayed activities in real time, allowing users to watch the automated tester at work.
The team later integrated all the components, ensuring the attacker communicated effectively with the victim AI systems and that the automated testing loop operated smoothly before polishing the final product.
They recorded a demo video and submitted the project before the 9 a.m. deadline on November 23, 2025.
During the afternoon judging session, the team delivered a pitch highlighting what they described as a major unaddressed market shift.
They argued that while the explosion of AI agents in 2025 has seen enterprises deploy them to manage everything from infrastructure to sensitive tasks such as financial analysis and customer support, small and medium-sized businesses remain vulnerable because they cannot afford to test such systems for security flaws.
Unlike major technology companies, SMEs lack the resources for dedicated AI security teams. Xploit, the team said, directly addresses this gap by positioning itself within the rapidly growing continuous automated red-teaming market projected to rise from $495 million in 2024 to $4.9 billion by 2032.
According to the developers, Xploit democratises AI safety by enabling businesses to automatically test and secure AI agents before deployment.
The judges were reportedly impressed enough to request access to the team’s code and development history to verify that the project had indeed been built during the hackathon weekend.
The team eventually won the “New Project Track” award and received $1,500 in prize money.
“What made the achievement particularly remarkable,” according to Kamdi Okeke, “wasn’t just that we built it over a weekend; it was that, competing among more than 100 of Philadelphia’s most driven creators, we built Xploit in less than a day of actual development time, transforming an abstract idea into a polished working prototype through focused collaboration and strategic planning.”
Speaking further, Okeke said another milestone came during UEV’s Venture Building Weekend held in Philadelphia from March 12 to 14, 2026.
“The mentorship and feedback we received from industry operators helped sharpen how we think about the problem and where our approach fits in the market,” he said.
United Effects Ventures (UEV), a Philadelphia-based pre-seed venture studio, organised the competitive hackathon focused on problem validation and go-to-market strategy.
After a gruelling 48-hour sprint, Xploit emerged tops, outperforming 15 competing teams, earning a cash award and two advisory sessions with UEV partners.
According to the team, the most significant outcome was the validation from industry experts, who endorsed Xploit’s focus on continuous red-teaming as an effective approach for identifying vulnerabilities in AI-powered products.
“Mentors at the hackathon validated both the team’s identification of the problem — the growing security risks posed by AI agents operating autonomously in enterprise environments — and our approach of framing the product as a continuous red-teaming platform capable of supporting an ongoing service model,” Okeke added.
