OpenAI partners with Hacktron to test system vulnerabilities

A new security partnership

OpenAI has officially partnered with Hacktron to investigate and address vulnerabilities within its systems. This collaboration follows a period where Hacktron researchers identified and disclosed security flaws through OpenAI's existing bug bounty program. The move marks a shift from independent reporting to a more structured, joint effort between the two organizations.

Under this new arrangement, engineers from OpenAI and security researchers from Hacktron will work together to select specific priority areas for testing. This collaborative approach ensures that the most critical parts of the infrastructure receive focused attention. By combining internal engineering knowledge with external security expertise, the companies aim to create a more robust defense against cyber threats.

As reported by EdTech Innovation Hub, Hacktron will also be granted expanded access to OpenAI's frontier models through a program called Daybreak Red. This access is intended to help researchers better understand the capabilities and risks of high-level AI. The partnership highlights a growing trend of using advanced models to secure the very systems they inhabit.

Testing connected systems

The primary focus of this partnership is the way OpenAI's systems interact with third-party services. As AI agents become more capable of connecting different tools and platforms, the complexity of the security landscape grows. OpenAI noted that these connections create new behaviors that require specialized testing to ensure they remain secure.

Hacktron specializes in building AI agents for tasks like code review, penetration testing, and red teaming. Red teaming is a security practice where professionals simulate attacks to find weaknesses before malicious actors can exploit them. Their previous research has specifically examined how applications, cloud infrastructure, and identity systems interact with one another.

The goal is to identify vulnerabilities that occur at the boundaries between these interconnected systems. Because a single flaw in one service can sometimes be linked to a flaw in another, testing these connections is vital. The partnership aims to uncover these complex, multi-step weaknesses that might be missed by looking at individual components in isolation.

Lessons from past research

The importance of this work was underscored by a previous account published by Hacktron regarding their research on OpenAI. In that instance, the researchers demonstrated how they could chain two different bugs together to reach the accounts of OpenAI employees. They achieved this access in under 72 hours using a harmless pull request to show the potential risk.

This specific example illustrates why testing the interaction between systems is so critical for modern security. It shows that even small, seemingly minor bugs can be combined to create a significant security breach. By focusing on these

Demonstrating chained vulnerabilities

Hacktron previously published a detailed account of their research into OpenAI's security. Their findings showed that it was possible to chain two separate bugs together to gain access to the ChatGPT and Codex accounts of OpenAI employees. This process was completed in under 72 hours, which served as a significant demonstration of how interconnected flaws can be exploited.

The researchers used a harmless pull request to show how they could reach these accounts. This specific case highlighted the danger of "bug chaining," where a minor vulnerability in one area is used as a stepping stone to reach a more sensitive target. The current partnership aims to proactively find these types of multi-step paths before they can be used by malicious actors.

Global recognition in edtech

In a separate announcement, the ETIH Innovation Awards 2027 have officially opened for new entries. The awards feature 26 different categories designed to recognize companies, institutions, projects, and individuals who are making a measurable impact on education technology. These global awards cover a wide range of topics, including AI in education, digital learning, and skills and workforce development.

The competition also highlights initiatives in higher education, inclusion, education operations, and evidence and impact. Entries are being managed through a new online platform and are open to participants from all over the world. Interested parties can explore the specific categories and judging criteria or read the "How to Enter" guide on the official platform to begin their submission.