Inside the alternative playbook to AI regulation

The U.S. Department of Commerce has approved the public release of OpenAI’s GPT-5.6 model, marking a major shift in the administration’s approach to regulating advanced artificial intelligence. According to a source close to the situation, the decision was made after extensive discussions between OpenAI and federal officials, culminating in the approval for a broad launch. OpenAI announced late Tuesday that its flagship model, Sol, along with lower-tier versions Terra and Luna, will be released publicly on Thursday. This follows a delayed rollout initially requested by the Trump administration, which had previously restricted early access to government-sanctioned entities. The approval for the wide release of GPT-5.6 comes after OpenAI conducted additional testing under the supervision of the Center for AI Standards and Innovation, a division within the Department of Commerce. Technical experts from the company spent several days in Washington, addressing potential concerns raised by government officials. These discussions were part of ongoing negotiations aimed at balancing innovation with regulatory oversight. The Trump administration had previously urged a staggered release of GPT-5.6, citing concerns over the risks associated with deploying highly capable AI systems. OpenAI expressed dissatisfaction with the staggered rollout, stating it was not the company’s preferred method for launching new models. This regulatory approach reflects broader challenges in governing emerging AI technologies. The Trump administration has emphasized a deregulatory stance, yet it has also taken steps to impose stricter controls on AI exports and usage. In June, the Commerce Department imposed restrictions on Anthropic’s Fable model, effectively withdrawing it from international markets. However, access to Fable was reinstated shortly after the administration revised its stance. Similarly, OpenAI’s GPT-5.6 faces scrutiny, though its public availability has been cleared by federal authorities. The situation highlights the evolving landscape of AI governance, where companies must navigate a patchwork of guidelines and enforcement mechanisms. Experts argue that the current regulatory environment lacks consistency and clarity. During the Biden administration, an executive order mandated that AI firms share safety testing results with the government, including assessments of vulnerabilities such as “jailbreaking”, a term used to describe instances where AI models can be manipulated to bypass built-in safeguards. This requirement was rescinded under the Trump administration, which sought to reduce bureaucratic hurdles. Without standardized criteria for evaluating safety risks, regulators and companies often find themselves in a reactive mode, adjusting policies based on individual cases rather than established frameworks. The lack of cohesive policy has created uncertainty for AI developers. Companies like OpenAI and Anthropic have faced delays and restrictions, even as they continue to innovate. OpenAI’s CEO, Sam Altman, described the recent approval process as “very productive,” acknowledging the effort made by both parties to reach a mutually acceptable outcome. However, critics point to systemic issues, including insufficient government funding and limited technical expertise. For instance, the Center for AI Standards and Innovation operates with an annual budget of only $15 million, despite needing $84 million to fully implement the Trump administration’s AI strategy. Such resource gaps hinder the ability to establish robust regulatory infrastructure. Congress has also lagged in passing comprehensive AI safety legislation, despite sustained bipartisan support and increasing public demand for oversight. Lawmakers like Rep. Josh Gottheimer have criticized the confusion surrounding the White House’s AI vetting process, arguing that it undermines both national interests and the competitiveness of American technology firms. Meanwhile, some states have taken initiative, implementing programs such as Oklahoma’s AI literacy training and Utah’s regulatory sandbox for testing AI under enhanced scrutiny. These localized efforts underscore the fragmented nature of U.S. AI governance. As OpenAI prepares to release GPT-5.6, the broader implications remain uncertain. While the immediate approval signals progress, the long-term effectiveness of the Trump administration’s regulatory approach will depend on how well it balances innovation with accountability. With more AI models likely to follow similar paths, the coming months will test the resilience of the current system, and whether it can evolve toward a more stable and transparent framework.