General James ‘Spider’ Marks: there’s a strategic risk hidden inside closed AI systems

Retired U.S. Army General James 'Spider' Marks warns that closed AI systems pose a strategic risk to military alliances as the performance gap between American and Chinese AI models narrows rapidly, citing Beijing-based Moonshot AI's Kimi K3 model achieving front-end coding capabilities rivaling Anthropic's Claude and OpenAI's ChatGPT. Marks argues that Anthropic's Fable 5 release, which included safety guardrails that blocked benign educational queries and selectively degraded responses on advanced AI development, exemplifies how private companies shape knowledge access, and he calls for a diversified 'best-of-breed' approach to AI procurement rather than dependence on a few closed systems.

Trust is the foundation of every military alliance. It is essential as partnership develops and ultimately is validated on the battlefield. It is becoming even more important as artificial intelligence assumes a larger role in national security. And it’s important for U.S. leaders to recognize that the performance gap between American and Chinese AI models is narrowing rapidly. In fact, the newly unveiled Kimi K3 model by Beijing-based startup Moonshot AI has achieved https://fortune.com/2026/07/17/xi-jinping-ai-cooperation-organization-shanghai/ front-end coding capabilities rivaling American titans like Anthropic’s Claude and OpenAI’s ChatGPT. The escalating friction between Washington and Silicon Valley’s frontier AI giants shows just how fragile trust between closed commercial ecosystems and government enterprises has become. Last month, Anthropic released Fable 5, its most powerful publicly available model. Arriving pre-equipped with stringent, safety-first guardrails https://urldefense.com/v3/ https:/www.nbcnews.com/tech/tech-news/anthropic-rankles-users-safety-first-fable-release-rcna349702 ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjflvOA08FI%24 designed to block or reroute dangerous requests, it was meant to prevent the general public from accessing its potent capabilities. Predictably, those restrictions cast a wide net. Reports https://urldefense.com/v3/ https:/www.nbcnews.com/tech/tech-news/anthropic-rankles-users-safety-first-fable-release-rcna349702 ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjflvOA08FI%24 indicate that Fable 5 declined to answer a range of benign educational questions in fields such as biology, chemistry, and mathematics – including basic https://urldefense.com/v3/ https:/x.com/cremieuxrecueil/status/2064433331970720187 ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjfl4H8gSZg%24 prompts https://urldefense.com/v3/ https:/x.com/cremieuxrecueil/status/2064433331970720187 ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjfl4H8gSZg%24 about mitochondria. Even worse, researchers accused Fable 5 of selectively degrading https://urldefense.com/v3/ https:/natolambert.substack.com/p/anthropic-walks-back-silently-nerfing ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjflNibL3N4%24 responses related to advanced AI development, providing less useful information on subjects that could aid competing developers and independent researchers. Whether intentional or not, the result is the same: a private company has been shaping access to knowledge according to its own interests. While the Commerce Department has since reversed course https://www.businessinsider.com/anthropic-mythos-5-us-restrictions-fable-5-openai-gpt-2026-6 , calls for government intervention have only increased, both from some concerned, well-intentioned policymakers and from Anthropic itself, seemingly looking to leverage regulatory capture to lock in its dominant position. The global AI landscape is evolving too quickly for static controls to secure U.S. leadership. National strength comes from cultivating a broad innovation base. No military commander would accept single point-of-failure dependence in weapon systems, a communications network, or an intelligence platform. That’s why U.S. defense leaders are choosing a “best-of-breed” https://federalnewsnetwork.com/commentary/2026/06/the-pentagon-is-rewriting-how-it-buys-ai-control-of-the-future-of-warfare/ approach to AI, partnering with multiple frontier technology companies, including providers of more transparent, adaptable open-weight models that Anthropic is campaigning against. This diffused strategy represents a welcome departure from the traditional model of rigid dependence on a small number of defense contractors. When power is concentrated among innovators – as it has been with American-built AI – select firms inevitably gain influence over how technology is developed, deployed, and used. Operating behind proprietary architectures, providers of closed AI models can modify algorithms, rewrite policies, or restrict access overnight without warning. They can dictate which lines of inquiry are permitted and which are discouraged. Open models offer a different path that’s crucial for America’s defenders. Rather than being granted access at the vendor’s discretion, they give governments and military organizations genuine operational control to pressure test proposed outcomes. Open-weight systems can be deployed on sovereign infrastructure, audited for security, tailored to mission requirements, and maintained independently of any commercial provider’s priorities. They also offer distinct geopolitical advantages. As JPMorganChase put it in a recent report https://www.jpmorganchase.com/content/dam/jpmorganchase/documents/center-for-geopolitics/cfg-us-china-report.pdf , China possesses important strengths in the AI race, including its robust support for open-source AI ecosystems, despite America’s leadership in frontier models, semiconductor design, and AI financing. Just last Friday at the 2026 World Artificial Intelligence Conference in Shanghai, Chinese President Xi Jinping used his keynote remarks https://fortune.com/2026/07/17/xi-jinping-ai-cooperation-organization-shanghai/ to position China as a champion of open-source artificial intelligence.The United States should not surrender that advantage. Rather, it should fast-track open-weight models built on democratic values that can be safely and securely used by America and its allies. Cybersecurity experts increasingly https://urldefense.com/v3/ https:/www.axios.com/2026/06/15/anthropic-fable-security-leaders-trump-admin ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjflhSZ2N4c%24 warn https://urldefense.com/v3/ https:/www.axios.com/2026/06/15/anthropic-fable-security-leaders-trump-admin ; K695gA XR4h0TG6Rk5xoqeDGOlr8sOj4F25aNLH8ILUuuDbdK6nzIvggrk30Hd83BWD5La0mG8h3KXODXVGHjflhSZ2N4c%24 that restricting access to advanced AI models hampers defenders’ ability to identify vulnerabilities, test systems, and develop countermeasures. In a world where perfect jailbreak resistance is unattainable, broader access to responsibly built models is more secure and narrows the risk of concentrated control. As AI becomes increasingly critical to communications, intelligence, and weapons systems, America needs AI infrastructure it can inspect, adapt, and control. Once frontier AI capabilities exist, they are unlikely to remain confined to a single company or country. If the U.S. does not lean into open-weight models, other countries will continue releasing them, while the domestic innovation ecosystem that underpins America’s long-term competitiveness erodes. National defense depends on trust, transparency, and resilience — and regulating America into a closed AI ecosystem provides none of the above. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune . Major General James “Spider” Marks is former Commanding General of the Army Intelligence Center, a member of the Military Intelligence Hall of Fame, and an on-air national security, military, and intelligence analyst for CNN.