After obtaining clearance from the US government, OpenAI will release its latest AI model GPT-5.6 on Thursday.

OpenAI will publicly release its most powerful model GPT-5.6 on Thursday. Previously, due to concerns that the powerful artificial intelligence (AI) model could be misused, the company postponed the release at the request of the U.S. government last month. Just a week ago, the U.S. government lifted restrictions on the release of Anthropic's two latest models, Fable and Mythos. Prior to this, the U.S. government had ordered the company to suspend access to these two models for national security reasons. According to reports, following additional government security testing under the new cutting-edge AI regulatory framework in the United States, the Department of Commerce has approved the full release of GPT-5.6. Previously, for national security reasons, the U.S. government had requested OpenAI to adopt a phased release strategy for GPT-5.6. Last month, the model was only allowed to be opened to a limited number of entities approved by the government. On Tuesday evening, OpenAI announced on the X platform that the company plans to simultaneously launch GPT-5.6 Sol, as well as the Terra and Luna models. Sol is the most advanced model OpenAI has developed to date, while Terra is a mid-range low-cost model, and Luna is the most cost-effective model. The U.S. government is tightening regulations on AI models. President Trump signed an executive order on June 2 strengthening the review of artificial intelligence models, requiring AI companies to provide the federal government with access to the model for up to 30 days before its release. Shortly thereafter, Anthropic's Fable 5 model was forced offline due to concerns about jailbreaking. Jailbreaking refers to an attack method that bypasses model security barriers using specific prompts. U.S. officials warned at the time that if Anthropic wanted to re-release Fable 5, it would have to address the security vulnerabilities. However, Anthropic believes that it is "almost impossible" to make any AI model completely resistant to jailbreaking attacks. The company also mentioned that there may be future universal jailbreaking methods that can unlock entire categories of harmful instruction behaviors at once. Prior to a major IPO, OpenAI seeks to relinquish ownership in exchange for regulatory easing It is worth noting that against the backdrop of soaring valuations in the AI industry and increasing political pressure, a game involving ownership distribution of AI giants and regulatory boundaries is continuing to ferment in Washington. There have been reports that OpenAI CEO Sam Altman and other executives are considering offering 5% ownership to the U.S. government in exchange for financial support and to resolve political obstacles. Based on OpenAI's record-breaking $85.2 billion post-money valuation after its funding round in March, this stake would be worth approximately $4.26 billion. Altman and others also proposed that major AI developers in the U.S. should allocate 5% of their ownership to a public investment vehicle, involving companies such as Anthropic, Google, Meta, and others. Last week, Trump stated that it is necessary to establish certain standards for AI, but regulation should not be too strict. "We do need some guardrails, but the fewer, the better." When directly asked about the reports of OpenAI negotiating to offer 5% ownership to the government, Trump avoided giving a direct answer. Currently, companies like OpenAI and Anthropic are preparing for their initial public offerings (IPOs). Whether relinquishing ownership can result in relaxed regulations and how subsequent regulatory boundaries are defined will affect the valuation of these AI companies.