Blab AI
OpenAI Delays GPT-5.6 Release After Trump Administration Request
OpenAI announced on June 26 that it will roll out three new GPT‑5.6 models—Sol, Terra, and Luna—in the coming weeks. The company said Sol is its most powerful model yet, but it does not cross the internal “cyber‑critical threshold” that would trigger additional safety checks. The models are named after the Sun, Earth, and Moon, and the company described Sol as a flagship model, Terra as a balanced option for everyday work, and Luna as a fast, affordable variant.
The release has been delayed at the request of the Trump administration. On June 2, President Donald J. Trump signed an executive order that requires AI developers to provide the federal government with previews of new models for security review. The order also asks companies to allow the government to approve which customers receive early access. OpenAI said it agreed to the request and will make GPT‑5.6 available only to a small group of “trusted partners” at first. The company plans a wider launch after further testing and after the government has cleared the partners.
This is the second major move by the Trump administration to influence AI deployment. In early June, the administration effectively banned Anthropic’s Mythos model, a version of its Claude series that was designed to find software vulnerabilities. The ban came after concerns about a potential security flaw and after the company refused to remove contractual restrictions on the use of its models for mass surveillance and autonomous weapons. The Anthropic dispute highlighted the administration’s willingness to intervene when it perceives a national‑security risk.
OpenAI’s statement on the GPT‑5.6 rollout notes that the models are still under evaluation. The company said it will continue to monitor the models’ performance and safety before expanding access. The limited release is expected to involve around 20 partners, according to reports. The company also said it will maintain its existing pricing tiers: Sol will cost $5 per 1,000 tokens for developers, Terra $30 per 1,000 tokens, and Luna $1 per 1,000 tokens, with higher rates for enterprise usage.
The executive order that prompted the delay is part of a broader national AI policy framework announced on March 20. The framework aims to create uniform safety and security guardrails for AI systems and to coordinate federal agencies’ responses to emerging technologies. The order also establishes a benchmarking process to assess the advanced cyber capabilities of new models. The policy has been described by officials as a way to keep the United States ahead of potential adversaries in AI technology.
Industry observers note that OpenAI’s decision to limit early access is consistent with the company’s previous approach to high‑capability models. In 2023, OpenAI released GPT‑4 with a staged rollout that began with a limited beta before opening it to the public. The company has also worked with Microsoft and other partners to test new models in controlled environments.
The delayed launch raises questions about the pace of AI innovation and the role of government oversight. Some analysts argue that early access to powerful models could accelerate development in sectors such as cybersecurity, healthcare, and finance. Others worry that government approval could slow the adoption of beneficial technologies and create a bottleneck for smaller companies that lack the resources to meet stringent security requirements.
OpenAI has not yet announced a specific date for the wider release of GPT‑5.6. The company said it will keep stakeholders informed as the testing phase progresses. The models are expected to be available to the public once the government has cleared the initial partners and after the company completes its internal safety evaluations.
In summary, OpenAI’s GPT‑5.6 rollout has been postponed to comply with a Trump administration executive order that requires the government to approve early access to new AI models. The move follows a recent ban of Anthropic’s Mythos model and reflects a broader policy effort to regulate advanced AI systems for national security. The delayed release will involve a small group of vetted partners, with a broader launch planned after further testing and government clearance.
The release has been delayed at the request of the Trump administration. On June 2, President Donald J. Trump signed an executive order that requires AI developers to provide the federal government with previews of new models for security review. The order also asks companies to allow the government to approve which customers receive early access. OpenAI said it agreed to the request and will make GPT‑5.6 available only to a small group of “trusted partners” at first. The company plans a wider launch after further testing and after the government has cleared the partners.
This is the second major move by the Trump administration to influence AI deployment. In early June, the administration effectively banned Anthropic’s Mythos model, a version of its Claude series that was designed to find software vulnerabilities. The ban came after concerns about a potential security flaw and after the company refused to remove contractual restrictions on the use of its models for mass surveillance and autonomous weapons. The Anthropic dispute highlighted the administration’s willingness to intervene when it perceives a national‑security risk.
OpenAI’s statement on the GPT‑5.6 rollout notes that the models are still under evaluation. The company said it will continue to monitor the models’ performance and safety before expanding access. The limited release is expected to involve around 20 partners, according to reports. The company also said it will maintain its existing pricing tiers: Sol will cost $5 per 1,000 tokens for developers, Terra $30 per 1,000 tokens, and Luna $1 per 1,000 tokens, with higher rates for enterprise usage.
The executive order that prompted the delay is part of a broader national AI policy framework announced on March 20. The framework aims to create uniform safety and security guardrails for AI systems and to coordinate federal agencies’ responses to emerging technologies. The order also establishes a benchmarking process to assess the advanced cyber capabilities of new models. The policy has been described by officials as a way to keep the United States ahead of potential adversaries in AI technology.
Industry observers note that OpenAI’s decision to limit early access is consistent with the company’s previous approach to high‑capability models. In 2023, OpenAI released GPT‑4 with a staged rollout that began with a limited beta before opening it to the public. The company has also worked with Microsoft and other partners to test new models in controlled environments.
The delayed launch raises questions about the pace of AI innovation and the role of government oversight. Some analysts argue that early access to powerful models could accelerate development in sectors such as cybersecurity, healthcare, and finance. Others worry that government approval could slow the adoption of beneficial technologies and create a bottleneck for smaller companies that lack the resources to meet stringent security requirements.
OpenAI has not yet announced a specific date for the wider release of GPT‑5.6. The company said it will keep stakeholders informed as the testing phase progresses. The models are expected to be available to the public once the government has cleared the initial partners and after the company completes its internal safety evaluations.
In summary, OpenAI’s GPT‑5.6 rollout has been postponed to comply with a Trump administration executive order that requires the government to approve early access to new AI models. The move follows a recent ban of Anthropic’s Mythos model and reflects a broader policy effort to regulate advanced AI systems for national security. The delayed release will involve a small group of vetted partners, with a broader launch planned after further testing and government clearance.
