Guardian: AI companies developing systems for war are already de facto defense contractors

Companies developing AI for combat operations should be called not just technology firms, but defense contractors. This thesis runs through a Guardian column arguing that the cost of weak AI-war regulation is already measured in human lives.

Hard Thesis

The text begins with an extremely direct formulation: from Gaza to Iran, the author writes, the same pattern repeats—precision weapons, deliberate blindness, and dead children. The main complaint is not with the rhetoric of precision itself, but with how it works in practice.

When a system promises surgical precision, political and legal accountability often becomes blurred: damage can be attributed to bad data, incomplete information, or model error. The author considers this the most dangerous effect of the current arms race in military AI tools.

The article's headline is framed as an accusation: if a company creates models and services that participate in target selection, threat recognition, or decisions about the use of force, it cannot be considered a mere software vendor. By the author's logic, this is already part of the defense supply chain, which means stricter rules for transparency, oversight, and accountability should apply. Otherwise, the technology brand becomes a convenient shield behind which the consequences of combat deployment can be hidden.

Logic of Fog

To explain this idea, the author refers to the Israeli 'fog procedure'—an unofficial rule by which soldiers at posts, in conditions of poor visibility, fire bursts into the darkness on the assumption that a threat might be lurking there. In the text, this practice is described not as random excess but as a principle: first comes blindness, then violence, and only afterward comes the explanation of why it was supposedly necessary.

AI, the author argues, does not cancel this logic but only makes it more systemic. Instead of a human shooting into the unknown, there is a model trained to operate on incomplete signals and probabilities. Formally, the solution appears more careful because it is accompanied by interfaces, labels, classification, and the language of 'precision.' But the essence, according to the column, does not change: the machine does not eliminate moral ambiguity but packages it in a technological procedure.

'Violence sanctioned by blindness.'

Where Responsibility Is Needed

From this thesis follows not only a moral but also a very practical dispute about regulation. If developers of military AI continue to call themselves simply AI firms, public discussion easily reduces to model quality, inference speed, and recognition accuracy. The author proposes a different angle: we need to discuss not just the technology but also its mode of application, the approval chain, data access, and the limits of permissible automation in matters of life and death.

Within this framework, the dispute over naming ceases to be semantics: the term determines which laws apply, who must disclose risks, and whom society can hold accountable for consequences.

• How exactly the system participates in target selection or threat assessment
• Who bears legal responsibility for model errors in combat operations
• What data and assumptions underlie the system's 'precise' conclusions
• Where the boundary lies between AI recommendation and a human's decision to pull the trigger
• What independent audits and constraints such products must undergo before deployment

What This Means

If we accept the logic of this column, the conversation about AI in war must shift from the comfortable image of a 'neutral model' to the question of the developer's status. When a product is embedded in the force application loop, it ceases to be merely a program. For states, this is an argument for direct regulation; for companies, it is a signal that they can no longer hide from responsibility for consequences behind the word 'technology.'