Analysis: Frontier AI Access Faces Economic and Security Constraints

An analysis published on 15 May 2026 argues that access to frontier artificial intelligence will become scarce and selective due to economic and security constraints. The article cites Anthropic’s limited release of its Mythos cybersecurity model and OpenAI’s restricted Daybreak initiative as evidence of a shift away from broad availability. Key factors driving this restriction include security risks such as misuse and model distillation, severe compute shortages, and potential US government intervention to prioritise national security and domestic economic interests. The author warns that these trends could create a geopolitical divide between "frontier haves and have-nots," recommending increased datacentre infrastructure and improved global cybersecurity as mitigation strategies.

In early April, Anthropic announced the development of Mythos, a leading cybersecurity model, restricting access to a select few companies, primarily US-based corporations. OpenAI has similarly committed to a limited release for its Daybreak initiative, which reportedly includes a model similar to Mythos (gpt-5.5-cyber). These moves dispel hopes that such restrictions were merely a fluke or marketing tactic, indicating a structural change in how top-tier AI capabilities are distributed. The analysis suggests that the mantra of abundant AI tokens driven by market pressures is being replaced by a reality of constrained supply.

The primary drivers for this restriction include security risks such as misuse and model distillation, severe compute shortages, and anticipated US government intervention. The US government is reportedly planning to intervene, potentially making restricted access a general rule for national security reasons. The author suggests the US may implement policies similar to the GAIN Act, giving American firms 'right of first refusal' for American-produced AI tokens. This intervention aims to protect national interests and prevent dangerous capabilities from falling into the hands of criminals or adversaries.

Compute shortages are exacerbating the situation, with leading developers facing significant challenges in balancing consumer subscriptions against real constraints on chip availability. Anthropic is reportedly seeking ad-hoc access deals with less well-utilised datacenters, including those of rival firm xAI, due to these crunches. The analysis notes that frontier capabilities have grown more expensive month-to-month, and efficiency curves will not mitigate the high marginal cost of providing access to new users or countries.

To mitigate these risks, the analysis recommends increased datacentre infrastructure and improved global cybersecurity. Non-US countries could secure frontier access by offering favourable terms, such as subsidised energy, to American hyperscalers for datacentre buildouts. The author warns that without these measures, the world could see a deepening divide where only a select few nations and firms have access to the most advanced AI, potentially leading to significant geopolitical and economic asymmetries.