Democratic senators have introduced a bill that would prohibit the military from using AI to spy on Americans or launch deadly attacks without human intervention.
The AI Guardrails Act, introduced Tuesday, was proposed by Michigan Sen. Elissa Slotkin, who infuriated the Trump administration by participating in a video encouraging military personnel to refuse “unlawful orders.”
Slotkin, a former CIA analyst, argues that the bill is necessary given the rapidly escalating use of AI in wars like the Iran conflict and the lack of regulation for rapidly evolving technology.
“If we were a more politically sane country today, we would have left-right restrictions on the use of AI,” she said at a recent Armed Services Committee hearing. “We don’t have any guidelines or laws for you. It’s not your fault. It’s our bipartisan responsibility.”
The bill would also ban AI-based decisions to launch a nuclear attack.
The Pentagon, whose negotiations with AI company Anthropic recently collapsed over ethical issues surrounding autonomous weapons, insists it is already complying with the law’s guardrails.
“The Department of the Army has no interest in using AI to conduct[illegal]mass surveillance of American citizens, nor does it want to use AI to develop autonomous weapons that operate without human involvement,” Pentagon official Sean Parnell said in a statement on X last month.
as independent person According to a report, the US military is making extensive use of AI in operations in Iran.
A key tool is Palantir’s Maven system, which, combined with Anthropic’s Claude large-scale language model, combs through intelligence, map data, and other systems to provide commanders with real-time information about the battle and help generate potential targets.
“These systems allow us to sift through vast amounts of data in seconds, allowing our leaders to cut through the noise and make smart decisions faster than our adversaries can react,” Adm. Brad Cooper, head of Central Command, said last week. “Humans will always make the final decisions about what to photograph, what not to photograph, and when to photograph it.…But with advanced AI tools, processes that previously took hours or even days can be reduced to seconds.”
Other major Silicon Valley companies such as OpenAI, Google, Elon Musk’s xAI, and Anduril all already supply or have contracts to supply various defense-related AI systems to the United States.
The US targeting process has come under intense scrutiny after an attack, likely by the US, destroyed an Iranian girls’ elementary school and killed at least 175 people.
In the run-up to the Iran war, the Pentagon negotiated contract terms with Anthropic on how to incorporate its AI into defense systems.
Anthropic claims that the Department of Defense wanted unrestricted use of its AI, and that the company wanted assurances that its products would not be used for mass surveillance or fully autonomous weapons. The Pentagon said it assured the company that the human model would be used only for lawful purposes.
Anthropic ultimately concluded that it “could not in good conscience” accept such terms, CEO Dario Amodei said last month, prompting Defense Secretary Pete Hegseth to declare the company a supply chain risk. The Trump administration has directed federal agencies to stop using Anthropic technology within six months.
Anthropic has filed a lawsuit challenging the risk designation, arguing that the administration is punishing the company for ideological reasons.
