Two Senate Democrats are drafting legislation to turn Anthropic’s controversial red lines into federal law. Sen. Adam Schiff (D-CA) is working on a bill to ensure humans make the ultimate decisions in matters of life and death involving AI, while Sen. Elissa Slotkin (D-MI) recently introduced the AI Guardrails Act, according to The Verge AI.
This comes after the Trump administration blacklisted Anthropic earlier this month, designating it a supply-chain risk after the company set limits on how the military could use its AI models. Anthropic has filed suit, accusing the government of violating its constitutional rights.
What Anthropic Refused to Do
Anthropic drew two hard lines with the Pentagon: no fully autonomous weapons and no mass domestic surveillance. That stance put it directly at odds with competitor OpenAI, which signed a deal with the Defense Department without those restrictions.
The administration’s response was swift and severe. By labeling Anthropic a supply-chain risk, the Pentagon effectively tried to cut the company out of government contracts entirely. Anthropic is now waiting to hear if a court will block that decision.
“The idea that they would therefore then try to turn around and kill the company, kill one of the preeminent leaders of AI is such a hostile, dictatorial kind of an act,” Schiff told The Verge AI. “They would set back America’s leadership in AI.”
What the Bills Would Do
Slotkin’s AI Guardrails Act would:
- Restrict the Defense Department from using AI to detonate a nuclear weapon
- Block AI-powered tracking of people or groups inside the US
- Require the Defense Secretary to notify Congress if “extraordinary circumstances” demand autonomous lethal weapons
Schiff’s bill is still being drafted, but covers similar ground. His office is talking with stakeholders and industry leaders before finalizing. Both bills draw on existing frameworks from the Biden administration.
The core principle is simple: humans stay in the loop. “Whenever a technology has the capability of taking a human life, there needs to be a human operator in the chain of command,” Schiff said. “We don’t want to delegate that kind of responsibility over life and death to an algorithm.”
That said, neither senator wants to ban military AI entirely. Schiff acknowledged AI can “tip and cue information for human operators” in real-time battlefield situations, helping distinguish civilian from military targets.
Why This Matters
This is the first serious legislative attempt to draw boundaries around military AI use in the US. Until now, those boundaries have been voluntary. Companies set their own terms, and the Pentagon decided which ones to accept.
That model clearly broke down. Anthropic tried to set terms and got blacklisted. OpenAI signed without restrictions and faced public backlash, then scrambled to claim it would insist on the same protections anyway.
Schiff isn’t buying the voluntary approach. “I would have a lot more confidence, frankly, if these were statutory requirements, than relying on the lawfulness of the Pentagon or the word of an AI CEO,” he said.
The Political Reality
Democrats hold the minority in both chambers, so passing anything requires Republican support. With midterms approaching, the window for new legislation keeps narrowing. Schiff is eyeing the National Defense Authorization Act (NDAA) as a potential vehicle, since it must pass annually.
“There’s certainly bipartisan support in the public for these kinds of limitations,” Schiff said, while acknowledging that some colleagues may resist anything that reads as implicit criticism of the administration.
The proposal could take another week or two to unveil. Whether it gains traction will depend on whether Republican lawmakers are willing to separate the policy question from the political one. The Verge AI has more details on the full legislative landscape.