HasItGot
Gaming

Leaked Anthropic Bill Would Limit AI Rules to Billion-Dollar Disasters

Anthropic has pitched draft legislation that only allows US intervention if an AI causes over 50 casualties or $1 billion in damage.

HasItGot Newsroom
Leaked Anthropic Bill Would Limit AI Rules to Billion-Dollar Disasters

Anthropic, the developer behind the Claude family of large language models, has submitted a private draft safety bill to United States lawmakers that establishes remarkably high thresholds before federal regulators can restrict artificial intelligence.

Under the terms of the confidential proposal, government authorities would only be permitted to step in against an AI system if it creates an imminent, predictable danger. Specifically, the model would need to threaten more than $1 billion in economic damages or risk killing or seriously harming at least 50 people.

Even if those casualty or financial limits are reached, the intervention powers would not apply universally. Federal agencies could only act if the hazard directly involves three specific scenarios:

  • The synthesis or deployment of biological weapons.
  • A digital attack conducted through cyberwarfare.
  • An artificial intelligence model successfully breaking free from developer constraints.

Industry Pushback and Current Harms

The draft measure, which was first uncovered by Politico and reported on by PC Gamer, shares notable similarities with recent legislative proposals introduced in California. A representative for Anthropic stated that the text embodies principles from the company's Advanced AI Framework, which staff have pitched to state and federal officials since June as practical statutory recommendations for Congress.

However, critics argue that these catastrophic benchmarks allow major tech firms to manufacture public consent for self-governance. By directing attention toward extreme, theoretical hazards, companies can avoid regulatory accountability for issues already occurring in the present day, including psychological impacts on consumers and disruptions across professional sectors.

The proposed legislation arrives during an era of significant friction surrounding safety inside top AI labs. Earlier this year, the head of safety at Anthropic resigned after warning that global security remains in jeopardy. Meanwhile, competitors face their own legal challenges; attorneys general across eight American states recently accused OpenAI of setting aside internal safety precautions in pursuit of commercial earnings.

Should lawmakers accept the framework drafted by Anthropic, regulatory agencies will remain sidelined from ordinary consumer and industrial harms. Unless a rogue system threatens mass casualties, bioweapons, or massive financial collapse, companies would be left to police their own development with minimal outside interference.

More gaming

All gaming