Artificial Intelligence Needs Guardrails

Commentary

Earlier this year, the CEO of the most powerful and popular artificial intelligence (AI) systems company warned that super-human intelligence will cause civilization-level damage unless there is a smart, speedy intervention.

“Humanity is about to be handed almost unimaginable power, and it is deeply unclear whether our social, political and technological systems possess the maturity to wield it,” Dario Amodei, CEO of Anthropic, said in a 38-page opinion piece about AI.

In his article, “The Adolescence of Technology: Confronting and Overcoming the Risks of Powerful AI,” Amodei expresses deep concern that the government, technology companies, and the public are vastly underestimating what could go wrong using AI. He says he wrote his essay to jar others, provoke a public debate, and detail the risks of AI.

“Humanity needs to wake up,” Amodei said.

He said his essay is an attempt to “jolt people awake.”

Amodei contends that he’s optimistic that humans will navigate this transition, but only if AI leaders and government are candid with people and take the threats more seriously than they do today. Amodei said that if we don’t address those challenges, we will face the stark reality of living in a “country of geniuses in a datacenter.”

Amodei lists several specific concerns surrounding AI and issues many cautions. He says time must be taken to carefully build AI systems that do not autonomously threaten humanity. Accordingly, it is critical for democratic nations to stay ahead of authoritarian nations and not be subjugated by them.

And here begins the rub.

Anthropic’s premier AI model, called Claude, has been widely used by the Defense Department to collect intelligence, identify targets, map out operations, and perform a host of other functions. In fact, Claude is currently being used in operations in Iran and Ukraine and was used in Venezuela to capture President Nicolás Maduro. Israel used it for operations in Gaza.

In its contract with the Defense Department, Anthropic said it had two restrictions. First, it said the government could not use its technologies for mass surveillance of U.S. citizens—figuring out where citizens have been, what they’ve visited, and what they’ve purchased. And second, it could not use Claude with autonomous weapons that kill humans without human involvement.

The administration, specifically President Donald Trump and Defense Secretary Pete Hegseth, disagreed with both restrictions. Anthropic argued it should be able to determine what’s off-limits for the AI systems it develops and sells to the government. The administration was contending that it wanted Anthropic’s tools “for all lawful purposes.”

Amodei said in a narrow set of cases, AI can undermine, rather than defend, democratic values. He said some AI uses are simply outside the bounds of what technology today can safely and reliably do. He says he doesn’t want to see his AI model used without human control in instances that could result in the killing of noncombatants or friendly troops by automated weapons. Nor does he want AI to be used to spy on Americans, which could violate deeply held values such as protecting privacy, ensuring freedom from illegal searches and seizures, and expressing dissent.

The fight between Anthropic and the administration has become heated and rather nasty, with Hegseth designating Anthropic a “supply chain risk,” a step never before imposed on a U.S. company. A government directive was subsequently issued that said no government contractor can do business with Anthropic.

The dispute between Anthropic and the government comes down to a matter of ethics and who sets the guardrails around (1) a technology that could mean an end to protecting the privacy of our citizens by enabling mass domestic surveillance, and (2) deferring to machines that can make life-or-death decisions on the battlefield without human involvement.

It could be reasonably argued that Anthropic has every right to go in a direction it chooses and to develop products that align with the company’s ethical parameters, core values, and moral positions.

On the other hand, there are those who would argue that our government should have full access to the best technology available, especially since China and other rival countries can ensure that companies in their countries comply with whatever their political leadership directs. Chinese companies do not request the government to put guardrails in place for their products. As The New York Times recently said, “So how China has used AI is the actual nightmare scenario that Anthropic is warning about.”

But in a democracy, guardrails are needed to control a growing technology such as AI. Clearly, some sort of guardrails are needed that help resolve the issues surrounding the use of AI. The need for these guardrails is now, and the responsibility for this task should rest with Congress. There is no time to kick the can down the road for this fast-moving technology. The time to establish policy or laws about how AI can be used is now. This is a far too important issue to let linger.

Originally published on The Baltimore Sun

Views expressed in this article are the opinions of the author and do not necessarily reflect the views of The Epoch Times.