OpenAI is preparing a limited-release strategy for its upcoming AI model, GPT-5.6, with access expected to be offered only to a select group of partners during the initial phase. The move comes amid reports of increased US government involvement in the evaluation of advanced AI systems before wider deployment.
The development was reported by The Information, which stated that the rollout approach has been shaped in consultation with US authorities. OpenAI has not publicly confirmed the details of the release plan.
According to the report, OpenAI CEO Sam Altman informed employees that government agencies would review and approve access on a customer-by-customer basis during the preview stage. The approach signals a more controlled distribution model compared to earlier model launches. The agencies cited in the report include the Office of the National Cyber Director and the Office of Science and Technology Policy. These bodies have reportedly pushed for closer oversight of advanced AI systems before they reach a broader user base.
The reported involvement reflects a broader change in the US administration’s approach to artificial intelligence. Earlier policy positions leaned towards lighter regulation. Recent actions, however, show a stronger push for federal review of powerful AI models. President Donald Trump recently signed an executive order that encourages voluntary submission of new AI models for government testing before public release. The policy aims to assess safety and misuse risks linked to increasingly capable systems.
One of the main concerns driving tighter oversight is the potential misuse of AI in cyber operations. Advanced models could assist in identifying software vulnerabilities at a speed that outpaces traditional human-led security research. Officials and researchers have repeatedly raised concerns that such capabilities could be misused if released without adequate safeguards or controlled access mechanisms.
OpenAI’s reported approach aligns with methods adopted by other AI developers. Earlier, Anthropic limited access to its cyber-focused model, Claude Mythos, through a restricted partner program. The company said the model carried an elevated risk if widely released. Such phased rollouts are increasingly used for high-capability systems where developers aim to balance innovation with safety considerations.
Apart from discussions on GPT-5.6, OpenAI recently introduced GPT-5.5 Instant, the default model powering ChatGPT. The company said the updated version offers improved understanding, stronger reasoning, and more natural interaction quality.OpenAI also announced Jalapeno, its first in-house AI chip designed for large language model workloads.
The development reflects the company’s broader push into hardware to support future AI systems. The direction of GPT-5.6 rollout suggests that access to advanced AI models may become more controlled as regulatory scrutiny and security concerns continue to grow.
Also Read: Anthropic Unveils Claude Fable 5 with Enterprise-Grade Reasoning and Security Guardrails