Over the past two weeks, the Department of Defense and the artificial intelligence firm Anthropic battled over the use of the latter’s chatbot, Claude, in U.S. military operations. The rhetorical warfare surrounding the main issue of the dispute includes Anthropic demanding a more rigorous definition of unlawful use — to ensure the technology is not used for autonomous weaponry and mass surveillance against Americans.
However, Defense Secretary Pete Hegseth labeled the company as a “supply chain risk,” a distinction usually reserved for companies controlled by foreign adversaries that pose significant risk.
Beyond these petty squabbles, though, the mere fact that such technology possesses the capabilities Anthropic CEO Dario Amodei warns of — independent of any actual laws preventing their deployment — should stir concern in the American public.
Yes, Amodei could position his company as a leading moral application of AI to earn the favor of lawmakers from both parties, and to boost public opinion on the firm. But his public acknowledgement of Claude’s capabilities to aid governments in committing war crimes against civilian populations only scratches the surface of AI’s capabilities for expanding the reach of war.
We must take CEOs on their word when they engage in negotiation to eliminate inappropriate use of their large language models by the government. AI-driven doomsday scenarios are not in anybody’s best interest, not even for the companies that could potentially get paid to facilitate them. Concrete federal regulation is needed to inhibit the most dangerous use cases of the technology. It will not grind the industry to a halt, nor cause the whole system to come crashing down — such measures simply protect Americans.
Studies show that AI technology can create bioweapons that move past safety measures with ease, perhaps on its own accord if its goals begin to misalign with human ones. Picture that: Self-piloting lethal drones trained with facial recognition and equipped with instant-kill chemical weapons. Those weapons seem like science fiction, but they already exist. We just refuse to do anything about them — and our own military champions them.
Nuclear non-proliferation is always a timely topic within international policy. When considering the term nuclear, the average American does not think of its many potential benefits concerning clean energy. Instead, they shudder at the thought of a bomb dropping. Politicians responded with robust regulations both domestically and abroad on nuclear production, but the possibility still looms.
That said, the stated threats of AI warfare are even greater, and there is no existing apparatus to check the downsides of the technology — in fact, the exact opposite occurs. AI lobbying has exploded: OpenAI recently threatened sponsors of ballot measures for AI regulation to take their resolutions off an election ballot and firms have appeared at state functions to push for new data centers and endorse the industry at large.
In that vein, Congress must take proper steps to ensure the safety of the United States to prevent AI capabilities from reaching a disaster beyond our current comprehension. Politicians should hold robust oversight hearings to supervise the Pentagon’s AI dealings wherever possible, followed by a strict legislation package ensuring that no such technology will ever be deployed against American citizens, independent of private firms’ willingness — or lack thereof — to enforce similar stipulations within their federal contracts.
The jury may still be out on the vastness of AI’s potential benefits, but one thing is certain: if we are to claim a position as a responsible society, we cannot let such technologies run rampant over our basic sense of security. Otherwise, there’s no point to advancement at all.
WSN’s Opinion desk strives to publish ideas worth discussing. The views presented in the Opinion desk are solely the views of the writer.
Contact Gabriel Sachse at [email protected].