Military AI Startups Forge Specialized Models for War Planning

The landscape of artificial intelligence development is witnessing a clear bifurcation, particularly concerning military applications. While frontier AI labs like Anthropic express reservations about providing unfettered access to their general-purpose models for defense, a new wave of startups is emerging, explicitly designing AI for military use cases. Smack Technologies exemplifies this trend, having recently secured $32 million in funding to develop models intended to surpass the capabilities of general-purpose LLMs like Claude in military operations planning and execution. This strategic focus highlights a pragmatic approach to defense technology, where specialized AI is tailored to the unique demands of warfare, rather than adapting broad-spectrum models.

Andy Markoff, Smack's CEO and a former US Marine Forces Special Operations commander, articulates a philosophy where the ethical deployment of military AI is entrusted to uniformed personnel, emphasizing adherence to the rules of war. This perspective contrasts sharply with the ethical stances of some general-purpose AI developers, as evidenced by the reported $200 million contract dispute between the Department of Defense and Anthropic, which stemmed from Anthropic's desire to restrict the use of its models in autonomous weapons. Markoff argues that general-purpose LLMs are inherently ill-suited for military tasks such as target identification or controlling physical hardware, primarily due to their lack of training on military-specific data and a human-level understanding of the physical world.

Smack's development methodology, which involves training models through extensive war game scenarios with expert analyst feedback—a process akin to Google's AlphaGo—underscores the need for highly specialized, domain-specific intelligence in military contexts. While Markoff clarifies that the current discussion is not about fully automating the "kill chain," the broader implications of such technology are significant. Rebecca Crootof, an authority on autonomous weapons law, confirms that the US and over 30 other nations already deploy weapon systems with varying degrees of autonomy. The future application of specialized AI, such as Smack's, is envisioned to automate the laborious aspects of mission planning, a task still predominantly manual. This shift could dramatically enhance the speed and efficiency of military decision-making and operational execution, but it simultaneously intensifies the ethical and regulatory challenges surrounding the integration of advanced AI into defense strategies. The emergence of these specialized firms signals a new era in military technology, where AI is not merely a tool but a foundational element of strategic and tactical capabilities.