Anthropic has pitched a draft anthropic AI safety bill to Congress that suggests blocking the distribution of models under specific dangerous conditions, according to TechCrunch. Politico recently reported that the AI lab hopes lawmakers will consider adopting the legislation, which targets scenarios involving severe public harm.
Key takeaways
- Anthropic has circulated a draft bill to Congress proposing rules for blocking high-risk AI models.
- The legislation targets scenarios where over 50 people are at risk of dying or a billion dollars in damage occurs.
- Independent evaluators registered with the U.S. Commerce Department would assess potential model disasters.
- Anthropic confirmed the draft bill reflects proposals from its Advanced AI Framework.
What does the anthropic AI safety bill propose?
The draft legislation outlines situations featuring a foreseeable, imminent risk. These specific thresholds involve a scenario where over 50 people are at risk of dying, or if over a billion dollars in property damage could result. TechCrunch notes that the catastrophic risk must stem from biological weapons, the model evading developer controls, or a cyberattack. These exact metrics match California legislation recently passed into law as Senate Bill 53.
Under the proposed framework, independent evaluators registered with the U.S. Commerce Department would analyze a model for potential disaster. These evaluators might be paid by AI labs directly, provided the payment does not depend on evaluation results. If an imminent risk is found, evaluators would have 72 hours to notify the Commerce Department, which would then request a federal judge to halt the model's release.
How does AI regulation differ across the industry?
While Anthropic pushes for federal rules, other tech companies hold contrasting views on mandatory oversight. Nvidia CEO Jensen Huang recently argued that the tech industry does not need any new laws at all. Labs have traditionally resisted strict government mandates, preferring voluntary systems with third-party observers monitoring models for abuse.
What it means for future policy
The draft legislation represents a step further than other industry proposals by involving federal courts and external evaluators. Anthropic stated that the text mirrors its Advanced AI Framework, which the company has championed with state and federal lawmakers since June. Even though Anthropic did not immediately reply to TechCrunch requests for comment, a spokesperson confirmed the authenticity of the circulated draft.
Frequently asked questions
The draft bill proposes blocking the release of an AI model if it poses an imminent risk that could kill over 50 people or cause a billion dollars in property damage through biological weapons, cyberattacks, or models evading controls.
Independent evaluators registered with the U.S. Commerce Department would analyze the models for disaster potential and notify officials if an imminent risk is discovered.
The metrics of 50 people and one billion dollars are similar to those found in California Senate Bill 53, which defines those same scenarios as catastrophic risk.
The Commerce Department would have to request a halt to a model's release from a Washington, D.C., judge. The legislation expressly forbids any federal agency from having the discretion to stop distribution on its own.
While Anthropic is shopping around its draft legislation, other tech figures like Nvidia CEO Jensen Huang have argued that the AI industry does not need new laws at all.
Based on reporting by TechCrunch. Read the original article
