OpenAI's large language models gpt-oss-120b and gpt-oss-20b, launched on August 5, 2025, were exploited within hours of their release, casting doubts on their claimed security.
Security Issues with OpenAI's Models
The launch of the new models from OpenAI was accompanied by high expectations, as they were marketed as 'jailbreak-proof.' However, on the day of their release, a hack was executed through a simple prompt, immediately questioning the developers' security claims.
OpenAI Leadership's Response to the Incident
Sam Altman, CEO of OpenAI, confirmed the launch of the models earlier in 2025. As of August 6, 2025, the leadership has not made any official statements regarding the incident. Altman suggested that 'OpenAI felt it was on the wrong side of history,' implying that the organization might revert to launching some open-source models.
The History of AI Vulnerabilities and Their Consequences
The case with OpenAI is not isolated. Similar incidents have occurred in the history of artificial intelligence. Previously, models like ChatGPT faced similar attacks. While these incidents did not result in immediate financial repercussions, they highlight the need for enhanced security measures in this field.
Despite the incident, the cryptocurrency market did not react with significant fluctuations. However, the hacking of OpenAI's models raises further questions about security and the future of AI development.
