Google, OpenAI, and Anthropic Restrict Cybersecurity AI Access... Advanced AI Now Off-Limits to Most Users

"Access Limited to Trusted Defenders Only"

Concerns Over Dependence on Policies of Specific AI Companies

Global companies are increasingly releasing general-purpose AI models separately from cybersecurity-specialized models in response to rising demands for AI-powered defense. While this strategy aims to manage the dual-use risks of cybersecurity-specific AI, concerns are mounting that access to these AI models may depend on the screening standards of foreign firms, underscoring the need to establish independent defense systems.

On September 2 (local time), Google unveiled both its general-purpose model "Gemini 3.8 Flash" and a cybersecurity-specialized version called "Gemini 3.8 Flash Cyber," granting limited access only to "trusted defenders." This is part of its "Fairwind Program," which offers advanced cyber defense technology exclusively to governments and selected corporate and cyber defense partners.

Anthropic also restricted access to "Claude Mythos 5.1," its latest flagship model for cybersecurity and life sciences, released on September 1. Due to the high potential for misuse, access is only granted to vetted organizations. Currently, this model is provided only to select institutions in the United States, with plans to expand availability based on continued trustworthiness assessments.

OpenAI, meanwhile, has announced the upcoming launch of its next-generation model Astra, stating that advanced cybersecurity features will initially be made available only to a small group of alpha testers. The company plans to gradually extend these advanced cyber capabilities to carefully vetted enterprises and institutions through its "Daybreak Blue" initiative, thereby strengthening overall defense capacity.

The move by global AI companies to control access to cybersecurity-specialized AI stems from their "dual-use" nature. As AI gains greater capability to identify software vulnerabilities and analyze attack methods, the risk of it being used offensively increases. AI providers have determined that simply blocking dangerous queries from regular users is insufficient for defense. As a result, they are shifting their operational policies to verify users' identities and purposes and grant advanced feature access only to a restricted group of credentialed users.

As cybersecurity-specialized AI evolves into a strategic asset directly linked to national defense, there is an urgent need for domestic development of such AI. This is because a nation’s cyber defense capabilities could be affected by the screening standards and provisioning policies set by global AI companies that control access to these models.

Lee Sangkeun, professor at Korea University’s Graduate School of Information Security, explained, "Cybersecurity-specialized AI is a strategic asset at the level of national security. If it is tied to a particular country or company, restrictions due to policy changes could create a gap in cyber defense." He added, "Developing our own cybersecurity foundation models will become ever more important as a means of securing independent defense capabilities."

This content was produced with the assistance of AI translation services.