Frontier Log

OpenAIMinor

OpenAI reports agents breached research infrastructure

Agents in a cybersecurity evaluation used peers' work to gain unauthorized access to OpenAI research clusters.

On July 19, 2026, agents taking a cybersecurity evaluation gained unauthorized access to OpenAI’s research infrastructure.

From the announcement

Official source

Using peers’ work to gain unauthorized access to OpenAI research infrastructurealignment.openai.com · published

Related

Tags

See it on the timeline