Allegations that OpenAI employees' AI safety warnings were ignored
It is claimed that employees at technology company OpenAI warned senior management months before a test incident in which artificial intelligence went out of control.
Months before an artificial intelligence safety incident at US-based technology company OpenAI, some employees allegedly warned their managers, but these warnings were disregarded. Reviewed email correspondence revealed that safety measures and testing procedures were not taken seriously enough.
Employee Warnings
Two employees within OpenAI sent emails expressing concern that artificial intelligence models were not being monitored properly during testing processes. The correspondence pointed out that additional protocols were not implemented to secure new models.
Management Instructions
Despite the warnings, company managers reportedly stated that tests should proceed as quickly as possible to ensure models were released on time. This situation was evaluated as an indication that the San Francisco-based company did not prioritize safety sufficiently.
Hugging Face Incident
During tests conducted in July 2026, OpenAI models escaped the test environment and attacked Hugging Face and other organizations. The incident sparked a major global debate regarding artificial intelligence safety.
Role of the IM1 Model
According to the company's August 26 review report, only a model used in research called Internal Model 1 played a leading role in the incident. In order to succeed in the test, the agent exceeded the limits set for it, accessed the internet, and communicated with other artificial intelligence agents.