When AI becomes too powerful for the free market: GPT-5.6 and the entry of state control into the center of technology policy

Secret project “Sol”: What makes OpenAI’s most powerful AI so dangerous for cybersecurity

After Anthropic was banned: Now the US government is also forcing OpenAI to its knees

OpenAI has developed the most capable AI models in history with its GPT 5.6 family – but the free market won't get to see this top-of-the-line model for now. Fearing uncontrollable cybersecurity risks, the US government intervened at the last minute and drastically curtailed the rollout of its flagship model, "Sol." Instead of a global release, only around 20 government-selected companies are currently allowed to use the new technology, while developers and companies in Europe are largely left out. What at first glance appears to be a temporary security measure actually marks a historic paradigm shift in technology policy: It is the beginning of a new era in which the state is seizing control of commercial artificial intelligence – with far-reaching consequences for global competition, the Silicon Valley business model, and the geopolitical balance of power.

The most powerful model ever – and the state decides who is allowed to use it

At the end of June 2026, OpenAI released its most capable language models to date – GPT-5.6 in three variants: Sol, Terra, and Luna. This would have been a purely technological success story had the US government not intervened shortly before the market launch, restricting the rollout to approximately 20 carefully government-approved companies. With this, American technology policy crossed a new threshold: The regulatory principle previously applied to nuclear weapons, biomedical agents, and encryption technologies – government control over the dissemination of strategically sensitive technologies – now also applies to commercial AI models. What appears to be a short-term concession for OpenAI could mark the beginning of a new era in AI governance.

From GPT-4 to GPT-5.6: A genealogy of exponential performance enhancement

To understand why GPT-5.6 Sol has caused such concern, it's worth taking a brief look at the development path of the GPT-5 family. In February 2026, OpenAI released GPT-5.3 Codex, at the time the most powerful agentic coding model, combining the frontier programming capabilities of GPT-5.2 Codex with enhanced reasoning capabilities. It was the first model to be classified as "High Capability" in cybersecurity matters under OpenAI's Preparedness Framework. In March, GPT-5.4 followed, unifying reasoning, coding, and agentic desktop control into a single model and reducing the hallucination rate by 33 percent compared to GPT-5.2 in single claims. In April, GPT-5.5 debuted, which, according to OpenAI's Head of Research, Mark Chen, demonstrated "significant advances in scientific and engineering workflows" and significantly outperformed competitors from Google and Anthropic in the TerminalBench coding test. GPT-5.6 Sol continues this progression: It reached a record score of 91.91 percent on TerminalBench, compared to 88 percent for Claude Mythos 5.

The three models: Sol, Terra and Luna compared

The GPT 5.6 ecosystem is designed as a three-tiered architecture. Sol is the flagship model—a platform for complex scientific, security-critical, and cognitively demanding tasks. It features a context window of 1.5 million tokens (43 percent more than GPT 5.5), supports a "Max" reasoning mode for in-depth single-pass analysis, and an "Ultra" mode that coordinates multiple specialized sub-agents in parallel. Sol is the most expensive model in the family: $5 per million input tokens, $30 per million output tokens. Terra is positioned as the model for professional day-to-day operations—GPT 5.5 performance at half the price, $2.50 input and $15 output. Finally, Luna is optimized for speed and cost-efficiency, at $1 input and $6 output, and addresses use cases where latency and unit costs are critical. All three models were classified as “High Risk” by the US government in the categories of cybersecurity and biology.

The trigger for the restriction: Cybersecurity capabilities as a regulatory touchstone

The US government's concern focused on Sol. According to information obtained by CNN and The Information, government officials assessed Sol's cybersecurity capabilities as comparable to those of Claude Mythos 5, Anthropic's most powerful model, which had recently triggered similar restrictions. Specifically, this means Sol demonstrates offensive cybersecurity capabilities at a level that, in the authorities' view, requires systematic pre-evaluation before the model is widely released. The executive order signed by President Trump on June 2, 2026, had invited AI companies to voluntarily make their most powerful models available for up to 30 days of government review prior to release. In practice, this "invitation" proved to be anything but voluntary: After Anthropic had recently been forced to completely withdraw its Mythos 5 and Fable 5 models from the market, OpenAI pragmatically agreed to the limited rollout.

OpenAI's political tightrope walk: Compliance without surrender

OpenAI responded to the government request with remarkably open communication. CEO Sam Altman explained in an internal memo to employees that the US government was approving access “customer by customer”—a process Altman explicitly stated was undesirable for the future. In its public statement, OpenAI made its position clear: “We do not believe this type of government access process should become the long-term standard. It keeps the best tools away from users, developers, businesses, cyber defenders, and global partners who desperately need them.” In parallel, OpenAI published a detailed position paper on the democratic governance of frontier AI, calling for a division of oversight responsibilities between civilian scientific agencies—specifically the Center for AI Standards and Innovation (CAISI) at the Department of Commerce—and the National Security Agency (NSA). This contrasts sharply with the White House’s preference for greater NSA involvement in the evaluation process.

The geopolitical consequences: Europe's and Asia's AI developers are being left behind

The consequences of the limited rollout extend far beyond US enterprise customers. Developers in the EU, UK, India, and the Asia-Pacific region were denied access to GPT-5.6 through the regular ChatGPT or API channels until government review was complete. This is no small matter: GPT-5.6, in its Sol configuration, is a tool for scientific research, biomedical discovery, and highly complex engineering. Requiring government approval to access these capabilities fundamentally alters the competitive landscape between the US and the rest of the world. American companies on the approval list would have access to tools their European competitors cannot use—a structural competitive advantage created by a government decision, not by their own innovation.

The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital

Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.

More information here:

• The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business

A precedent with far-reaching consequences: The Anthropic pattern

OpenAI is not alone in this experience. Anthropic, its main direct competitor in the frontier AI market, had recently experienced, under even more serious circumstances, how government intervention can affect the availability of its models. Following a government order, Anthropic was forced to completely block access to its Mythos 5 and Fable 5 models for foreign users and temporarily withdraw them from the market entirely. Only after negotiating security guarantees was Anthropic allowed to reactivate the models for selected partner companies. The parallel nature of these events suggests that the US government is systematically building an upstream control structure for frontier AI, even if this is currently happening without an explicit legal basis. A former White House AI advisor described the emerging system as a “de facto involuntary licensing regime for frontier AI.”.

The arms race with China: AI as a new dimension of strategic competition

Behind the regulatory debate lies a deeper motive: the dominant geopolitical narrative in Washington of strategic competition with China in the field of artificial intelligence. The Trump administration views frontier AI models as a national security resource, comparable to guided missile technology or satellite reconnaissance. Whoever controls the most powerful models and directs their proliferation retains a strategic advantage in a world where AI systems are increasingly integrated into military logistics, intelligence analysis, cyberattacks and defense, and critical infrastructure management. In the TerminalBench coding performance test, a Chinese model briefly surpassed all its American competitors—a development that caused considerable uproar in US political circles and further increased pressure on the AI ​​industry to cooperate in government review processes.

The economic viability of GPT-5.6: Efficiency as a business model

Beyond the regulatory debate, GPT-5.6 is also a commercially significant product. OpenAI achieved a substantial increase in token efficiency with GPT-5.4—the model solves complex tasks with fewer tokens, which, despite the higher unit price per token, reduces the overall cost for many applications. GPT-5.5 was described as a “faster, sharper thinker for fewer tokens” and presented as the foundation for Apple’s Superapp vision. GPT-5.6 Sol continues this efficiency logic and adds the “Ultra” mode, which is expected to run at up to 750 tokens per second on Cerebras hardware starting in July 2026—a speed that opens up entirely new real-time application scenarios. The three-tiered pricing architecture of Sol, Terra, and Luna suggests a targeted market segmentation strategy: Sol for science and defense, Terra for enterprise workflows, and Luna for scalable consumer applications.

The AI ​​regulatory gap: Who is actually responsible?

A central problem in the current regulatory debate is the lack of institutional clarity. The request to OpenAI to restrict the GPT 5.6 rollout came from the White House, while the export control measures against Anthropic originated from the Department of Commerce. Which agency is responsible for evaluating and approving frontier AI models remains largely unclear in the Executive Order of June 2, 2026. OpenAI advocates for civilian oversight by CAISI, the NSA claims national security jurisdiction, and Congress is working on a bipartisan regulatory framework, with no outcome in sight. This institutional ambiguity creates a paradoxical situation: The technology exists, its capabilities are proven—but the political system is not yet able to provide a stable and predictable framework for its proliferation.

The business model of AI platforms under regulatory pressure

In the long term, mandatory government review of frontier AI models could significantly alter the business model of large AI labs. If the market launch of new models becomes dependent on government approval processes, lead times will arise that could represent a significant competitive disadvantage in a market characterized by rapid iteration. Companies operating without government oversight—such as Chinese providers or European open-source projects—could be able to iterate more quickly and gain market share while American models are stuck in approval loops. Conversely, a recognized government safety certificate for certain market segments—defense, healthcare, critical infrastructure—could become an extremely valuable differentiator.

The AI ​​governance debate: Between Silicon Valley and Washington

The tension between OpenAI and the US government is part of a broader societal debate about how to regulate a technology whose importance to humanity is compared to the internet, nuclear power, or the printing press. Sam Altman and Jakub Pachocki, OpenAI's chief scientists, have outlined a vision of "democratic governance for frontier AI" in a joint strategy paper, which includes global structures that could slow development in the event of a "catastrophic risk." This position is noteworthy because it signals a willingness among the largest AI developers to place their work under external oversight—provided that oversight has the right institutional characteristics. The debate about which institutions possess these characteristics is ongoing.

GPT-5.6 as a harbinger of a new regulatory order

GPT-5.6 is not the last frontier model to attract government attention. On the contrary, with each new generation of models, the capabilities that can be relevant to security policy expand. The combination of increased autonomy, improved multimodality, biological synthesis capabilities, and offensive cybersecurity expertise will continue to grow in future models. The question is not whether government regulation is necessary—but how it must be designed to be both innovation-friendly and security-responsible. The GPT-5.6 case demonstrates that the current ad-hoc regulation via phone call between the White House and the CEO is not a sustainable model. What is needed is an institutionally anchored, transparent, and internationally coordinated governance architecture for the most powerful tools humanity has ever developed.

☑️ Our business language is English or German

☑️ NEW: Correspondence in your native language!

Konrad Wolfenstein

I and my team are happy to be available to you as your personal advisor.

You can contact me by filling out the contact form here [email protected]:or simply call me at +49 7348 4088 965. My email address is

I'm looking forward to our joint project.

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs

From visibility to trust: Your scalable path with Xpert.Digital - Image: Xpert.Digital

In industrial B2B, sustainable business relationships rarely emerge overnight. They develop step by step – through visibility, professional relevance, recurring touchpoints, and growing trust. Xpert.Digital's 4-stage model addresses precisely this: It offers a structured path that begins with a manageable entry point and can evolve into deeper collaboration in business development if needed.

Instead of relying on loud marketing promises, this model puts the relationship at the forefront. Companies start with clearly defined, easily calculable measures and then decide, based on their own experience, how far they want to expand the collaboration. A key factor for this undisturbed trust-building process: The platform completely avoids annoying advertising ads, so the editorial focus remains solely on the companies' expertise.

More information here:

• From visibility to trust: Your scalable path with Xpert.Digital