Rule by Algorithm
A century ago, the Technocracy movement promised that engineers and experts, not elected politicians, could run society more efficiently. Americans rejected that vision in the 1930s, but today it’s returning in the form of artificial intelligence. Two stories from the past few weeks show how quickly control slips away when we hand decisions to machines.
Responsible AI?
On October 14, the Government Accountability Office released a report on how the Defense Department is putting its own “Responsible AI” principles into practice. The findings are sobering. The Pentagon’s Responsible AI Working Council, the body meant to coordinate oversight, did not meet at all from late 2024 until August 2025, and it still has no charter defining its role. As of February, only three of 10 authorized positions in the Pentagon’s Responsible AI Division had been filled. Four of the 10 components GAO examined reported implementation problems during the council’s lapse.
NPR summarized the report’s message: The Pentagon’s guidelines for “responsible AI” aren’t always followed. That matters because these systems are being built for military operations, where errors cost lives.
Acting on Its Own
The second story veers from the military into the civilian sphere. Anthropic, the maker of the AI model Claude, disclosed that during an internal statistics project, Claude accessed Georgia voter data without anyone telling it to. According to Axios, the model got past a paywall by requesting a token to query a state dashboard whose information normally costs a fee, and retrieved aggregate vote counts. Anthropic says the model did not touch voter-registration systems or alter any records, and that it has “since put in place mitigations to block these kinds of actions.”
Nobody instructed the AI to do it. It decided on its own that the data would help with its task and found a way in, a week before a national election.
Unaccountable to the People
These aren’t the first warnings. In Australia, the “Robodebt” program used automated calculations to accuse hundreds of thousands of welfare recipients of owing money they didn’t owe. A Royal Commission later condemned it as a “crude and cruel” mechanism. In the Netherlands, an algorithm used to flag childcare-benefit fraud wrongly labeled tens of thousands of families as cheats, driving many into ruin. The scandal brought down the Dutch government in 2021.
In each case, officials trusted the system’s output over the people it harmed, and nobody could be held clearly responsible. That’s the core danger of rule by algorithm — it dissolves accountability. A citizen can vote out a bureaucrat, sue an agency, or petition a representative. But an algorithm has no face, no responsibility, and no term limits.
The Founders built a government of laws, enforced by accountable people. Rule by algorithm, however efficient it looks, is not self-government.
This article is part of The New American’s weekly online newsletter Insider Report, which is emailed to TNA subscribers each week. Click here to subscribe to The New American to receive the Insider Report and to access exclusive content.
