OpenAI announced on November 6 that its engineers will work with Hacktron researchers to find vulnerabilities in its systems. The teams will choose testing priorities together, focusing on how those systems connect to external services. [1]
What changes
Hacktron had already reported vulnerabilities to OpenAI through its bug bounty program. Under the agreement, the teams can plan tests together, and Hacktron will get broader access to OpenAI's frontier models through Daybreak Red. [1]
Context
The work focuses on connections between applications, cloud infrastructure, and identity systems. The announcement leaves the specific testing targets open. [1]
What to check in practice
To understand the scope, imagine an application that connects an assistant to another provider’s sign-in system. A useful test would trace which permissions survive when an account changes role. That example explains why testing system boundaries matters. The announcement gives neither a list of planned tests nor a service rollout by country.