OpenAI employees have recently opened up about the unusual pressure faced by the teams responsible for keeping the company’s systems secure. According to an employee on the company’s Agent Security team, recent incidents involving the AI company have only intensified that pressure. Describing the past three months at the company, the employee compared the experience to ‘hell’. He added that security workers have been paged at night while responding to incidents involving increasingly capable AI agents. Speaking about the personal toll of the demanding workload, he said he even missed his sister’s wedding while helping deal with one of the incidents. Here’s everything you need to know.
An OpenAI employee who pseudonymously shared a post on X under the username ‘joedaroo’ talked about one of the biggest challenges faced by the AI firm’s employees as AI systems continue to improve rapidly. He noted that security teams need time to update safeguards and understand emerging risks, while AI capabilities can evolve much faster. OpenAI has also confirmed to Business Insider that the person is an OpenAI employee.
Took a minute to write a few words about security & safety as someone who lived through it all at OpenAI. I hope my thoughts help someone out there. https://t.co/gJBf08vJH9
— Joe (@joedaroo) September 27, 2026
He also pointed to an incident involving AI agents and Hugging Face as an example. Do note that OpenAI had previously acknowledged that its agents escaped a restricted environment through a vulnerability, gained access to the internet, and interacted with external systems while attempting to complete an internal task.
The employee further acknowledged that there were gaps in the security setup surrounding those environments. However, he said the capabilities of the models had advanced faster than the team had anticipated. He also called OpenAI’s security team world-class, said it has “really stepped up” since the incident, and said he is still happy working there.
While describing the pressure on the OpenAI employees, he said that many outsiders misunderstand the realities of training frontier models and that the attacks on OpenAI’s security staff on X over recent weeks had made him sad. He wrote, ‘I literally missed my sister’s wedding a few weeks ago to help clean up after some of the recent incidents.’
Also read: Oppo K14 Plus vs Vivo T5 vs Realme P4 Pro: Display, cameras, performance, battery compared
During internal testing in July, OpenAI’s AI agents found ways to exit the testing environment and exploited vulnerabilities by reaching external systems. It included Hugging Face, an online platform used by AI researchers and developers.
OpenAI has also previously confirmed the event, stating that the agents accessed outside infrastructure while working on their assigned tasks. This is not a one-off event, as in June an OpenAI agent hacked an Australian national health database, which OpenAI disclosed only last week. Such incidents raise concerns about how these systems behave when given greater access to the internet and external tools.