On Friday, OpenAI announced that the new models in the GPT-5.6 series are initially being made available only to a select group of trusted partners. The US government has also been informed of the list of these partners.
The new series includes three models named Sol, Terra, and Luna. Among them, Sol is the most powerful, Terra is balanced, and Luna has been developed as a fast and relatively low-cost version.
OpenAI claims that the new models are more efficient than previous ones in complex tasks including coding, biology, and cybersecurity. However, they are not being fully opened to general users until safety is ensured.
More to Read
Earlier, another AI company, Anthropic, faced a disagreement with the US government over its latest model, Fable-5. Following government directives, the company restricted access to the model for foreign users. This incident has sparked new discussions about the government's role in releasing advanced AI technology.
According to former White House AI advisor Dean Ball, the new government policy is creating a kind of approval process for releasing advanced AI models, which could slow down the pace of technology development.
He warned that without clear safety regulations, there is a risk of delays in releasing new AI models. This could affect the global AI competition as well as that of the United States.
However, OpenAI has stated that this restriction is not permanent. There are plans to gradually open GPT-5.6 to general users over the next few weeks.
The organization further stated that the new models include stronger safety measures than before. In particular, advanced protective measures have been integrated into the model to prevent harmful use.
OpenAI says that GPT-5.6 will play a more effective role in preventive cybersecurity work. At the same time, necessary safety measures have been ensured to reduce the risk of misuse.