AppVolt AI news, read from the source

OpenAI asks outside hackers to find holes in its own systems

OpenAI is working with the security team Hacktron to test how its systems connect to outside services, and is giving it wider access to its newest models.

November 9, 2026 · 1 min read

Partnering with Hacktron to find vulnerabilities in OpenAI’s systems in white text on an abstract blue, pink, and lavender background.
Image: OpenAI

OpenAI said on November 6 that it is working with Hacktron, an outside security team, to find weak spots in its own systems. Engineers from both sides will pick the most important areas to test together.

The focus is on how OpenAI's systems connect to outside services. As part of the deal, OpenAI is giving Hacktron wider access to its newest AI models through what it calls Daybreak Red.

Hacktron builds AI agents that review code, test systems for weak spots, and play the attacker in practice drills. OpenAI says it first got to know the team through its bug bounty program, where outside researchers report flaws for a reward. It says it was impressed by the flaws Hacktron's researchers found and reported there.

Why the connections matter: AI agents now link services in new ways, OpenAI says. Testing those links more deeply should help it make its defenses stronger. Zayne Zhang, the head of Hacktron, said the work will hunt for complex flaws "at the boundaries between interconnected systems."

The deal is part of a wider push. Earlier the same week, OpenAI said researchers from a team called Cathedral will join its next security sprint. It also raised its bug bounty rewards. The goal is to give independent researchers more reason to chase flaws its own testing may have missed.

For a small business that links many apps together, it is a useful reminder. Check how your tools connect to each other, not just each tool on its own.

What this does not mean. It does not mean ChatGPT is unsafe to use, or that Hacktron found a flaw in your account. This is a testing partnership, not a report of a breach.
Primary source OpenAI · published November 6, 2026
Open it and check the claim yourself. That is the point.

One email a week

A digest of the week's verified stories. No daily flood, no tracking pixels, unsubscribe in one click.

We send the digest and nothing else.