What Happened?

Congress has not passed a law specifically governing how the Department of Defense (DOD) uses frontier artificial intelligence (AI) models or how reliable they are in autonomous weapons. ​

The Congressional Research Service (CRS) updated its report on Oct. 5, identifying potential congressional responses amid the Pentagon's dispute with Anthropic.

President Donald Trump directed federal agencies to stop using Anthropic's technology in February. Defense Secretary Pete Hegseth then ordered the Department of Defense (DOD) to cut Anthropic from its supply chain, giving the agency up to six months to transition away from its products.

A divided federal appeals court upheld that exclusion last month, finding sufficient evidence of a national-security supply-chain risk and rejecting Anthropic's constitutional claims.

Why Does it Matter to Me?

Lack of regulation over AI has become a political issue. With news stories of late of rogue AI agents and chatbots, many Americans from all sides of the political aisle are concerned about AI according a recent Pew Research Center survey.

The survey goes on to report that Americans are fairly divided on whether they trust the U.S. to regulate AI effectively.

Both Sides, Now

A Pentagon directive updated in January 2023, DOD Directive 3000.09, distinguishes between fully autonomous, human-supervised, and semiautonomous weapon systems. It requires appropriate levels of human judgment over the use of force but does not require a person to manually approve every individual engagement.

Congress has set some reporting requirements through recent defense authorization laws but has not passed legislation specifically restricting how frontier AI models may be used in weapons.

Anthropic argued it was willing to modify its usage policies for DOD but refused to authorize mass domestic surveillance or fully autonomous weapons. CEO Dario Amodei said "frontier AI systems are simply not reliable enough to power fully autonomous weapons" as part of Anthropic's red lines and public explanations regarding why the company refused to allow its Claude AI model to be used for fully autonomous weapon systems without human oversight.

Concerns about AI-assisted targeting are not hypothetical. A February 2024 U.S. airstrike killed 20-year-old student Abdul-Rahman al-Rawi; Airwars and The Independent reported it as a known case involving acknowledged AI-assisted targeting, and U.S. Central Command acknowledged an internal investigation. Airwars noted the incident does not establish that AI itself caused the death.

DOD is not publicly known to use Anthropic's Claude or other frontier AI models in autonomous weapons. The appeals court sided with DOD, and Anthropic said it was considering its options after the ruling.

What Happens Next?

Anthropic said it was considering its options after the ruling, meaning further court review remains possible. The exclusion stays in force unless a court overturns it.

AI Guardrails Act (S. 4113), a Senate bill, has been introduced but not enacted.

Congress can act through the annual defense authorization process, as the CRS report reviewed on Legis1 notes, but no action is scheduled as of now.

Will Congress set clear rules for when and how the military can use AI to make lethal decisions, or leave those choices to the Pentagon and the courts?

---

AllCiv makes it easy to stay politically informed and involved.

---

Spot something wrong? Report an issue with this article