The Pentagon signed a deal with Elon Musk’s xAI that would allow the Grok chatbot to be used on classified systems and for “all lawful use,” according to Axios.
To date, Anthropic’s Claude has been the only artificial intelligence model that the Pentagon has used for its most sensitive operations. However, a conflict recently arose between the Pentagon and Anthropic over whether the AI company would allow its model to be used for purposes involving autonomous weapons and mass surveillance of Americans.
Secretary of Defense Pete Hegseth and Dario Amodei, the Anthropic CEO, scheduled an in-person meeting to discuss the issues, per Axios.
According to The New York Times, the Pentagon also has been in talks with OpenAI and Google about using their AI models on classified systems. The Pentagon has insisted that those companies agree to the “all lawful use” standard, as well.
Experts have raised concerns about the risks posed by using AI for military operations. Diplo, a nonprofit organization, has pointed to issues such as black-box decision-making and bias.
“Black-box decisions refer to unexplainable outputs by an AI system,” the company explained. “For example, an AI may assign features to a target or calculate a score for suspect analysis without understandable logic for the system’s conclusion.”
If military officials are unable to understand the basis for AI decision-making, it would be difficult if not impossible to gauge the soundness or accuracy of those decisions.
Black-box decision-making is of particular concern in military applications, wherein split-second decisions can mean the difference between life and death.
Similarly, officials operating in the fog of war could have difficulty determining whether and to what extent an AI decision has been influenced by bias.
“Bias in AI systems is inevitable,” according to Diplo. “… In a military context, AI is usually trained on data from surveillance footage, behavioural patterns, and biometric databases, which can be skewed by profiling based on race, religion, or geography.”
Because of these risks, many AI experts believe it is essential for AI models to have strict guardrails. However, by applying its “all lawful use” standard, the Pentagon appears to be pushing back against such safeguards.
The widespread adoption of AI has impacted far more than just the military. For everyday consumers, electricity prices have spiked as the energy-hungry data centers that power AI models have increasingly strained the electrical grid.
Get TCD’s free newsletters for easy tips to save more, waste less, and make smarter choices — and earn up to $5,000 toward clean upgrades in TCD’s exclusive Rewards Club.
