Tell Congress: Stop Big Tech's 50-Death AI Safety Gamble

How many people should have to die before the government can stop a dangerous AI system? According to newly revealed legislation proposed by AI giant Anthropic, the answer could be more than 50.

Anthropic, the company behind Claude, claims to be a champion of AI safety. But behind closed doors, it's been pushing legislation that could give AI companies far more protection from government intervention than the people exposed to their technology.

Under Anthropic's proposal, federal officials seeking to block certain dangerous AI models would have to demonstrate an imminent threat of more than 50 deaths or serious injuries, or over $1 billion in damage, from catastrophic risks such as biological weapons or cyberattacks. That's a threshold for predicted harm, not actual deaths—but it could still prevent regulators from acting against serious threats before they reach that scale.

That's an outrageous standard to write into law. Why should regulators have to predict a catastrophe of that magnitude before they can take preventive action?

Tell Congress: Don't let AI companies write safety rules that put corporate interests ahead of human lives.

The proposed restrictions don't stop at arbitrary numbers. Anthropic also wants to shield certain AI safety records from public information requests and override some state AI protections. That could mean less transparency, fewer safeguards, and more power concentrated in the hands of the very companies we're supposed to be regulating.

The hypocrisy is staggering. Anthropic built its reputation on AI safety. Now it's asking Congress to make it harder for regulators to prevent catastrophic harm.

Dangerous AI systems could help criminals launch cyberattacks that shut down hospitals, disrupt power grids, or expose millions of Americans' private information. They could even help bad actors develop biological weapons. When the stakes are this high, regulators need the power to prevent catastrophe—not wait until the danger becomes impossible to ignore.

We wouldn't let pharmaceutical companies decide how many people must be harmed before regulators can stop a dangerous drug. Why should AI companies get to set the rules for their own technology?

Congress must draw the line now, before industry lobbyists turn this proposal into law.

Add your name: Tell Congress to reject Big Tech's dangerous AI safety loopholes and put public safety before corporate power.

Thanks for all that you do,
Matt from the Swarm