The White House has asked OpenAI to slow down the release of its new model, GPT 5.6, citing safety concerns. The company will share the model with a select group of close partners before making it available to the public, in what is expected to be a limited release lasting several weeks.
This decision comes as part of a broader effort by the Trump administration to exert greater control over the development and deployment of advanced AI models. In recent months, the government has pushed for federal oversight of new models, with President Trump signing an executive order directing certain companies to submit their models for testing and evaluation before releasing them publicly.
OpenAI's move is reminiscent of Anthropic's decision earlier this year to limit access to its powerful frontier cyber model, Claude Mythos. The company argued that the model was too powerful and could be misused by malicious actors if released widely. While some have questioned the legitimacy of this concern, others see it as a necessary precaution in an era where AI-powered tools are increasingly being used for nefarious purposes.
The use of advanced AI models has raised concerns about their potential misuse by cybercriminals. These models can write malware and execute ransomware attacks autonomously, making them a significant threat to organizations with complex software infrastructure. The government's decision to slow down the release of GPT 5.6 is likely an attempt to mitigate this risk and ensure that these powerful tools are not misused.
OpenAI has reportedly worked closely with government agencies on the upcoming release, including the Office of the National Cyber Director and the Office of Science and Technology Policy. The company's decision to limit access to its new model for now may be seen as a compromise between the need to make AI more widely available and the need to protect against potential misuse.
Cybercriminals have long used automated tools, but the age of generative AI has given them more digital ammunition than ever before. Large language models (LLMs) can write malware and execute entire ransomware attacks autonomously, posing a significant problem for organizations with complex software infrastructure. The specific concern with frontier cyber tools like Mythos is that they are capable of identifying and exploiting software vulnerabilities at speeds no human analyst could match.
The Trump administration's push for federal oversight of new models has sparked debate about the balance between innovation and regulation. While some see it as a necessary step to protect against potential misuse, others argue that it stifles progress in the field. OpenAI's decision to slow down the release of GPT 5.6 is likely an attempt to navigate this complex landscape and ensure that its powerful tools are used responsibly.
