Anthropic pitches AI safety bill to block risky models
Anthropic has circulated a draft AI safety bill proposing to block models if they risk killing 50+ people or causing $1 billion in damage.
Stock photo for illustration only, not from the actual event
- Anthropic pitches a new draft AI safety bill to US lawmakers
- Proposes blocking model releases under catastrophic risk scenarios
- Sets thresholds at 50 fatalities and one billion dollars in damage
- Aligns with Anthropic's previously stated Advanced AI Framework
Over the past year, intense concerns have surrounded the potential dangers of artificial intelligence, sparking continuous debates over proper regulatory paths. Now, Anthropic has put forward its own vision for how the government might oversee the industry.
Politico recently reported that the AI lab has been circulating a new draft bill regarding AI safety, hoping Congress might consider adopting it. The legislation outlines specific scenarios in which the release of an artificial intelligence model could be legally blocked from distribution.
The metrics introduced in the draft — 50 people and one billion dollars — mirror those found in California’s Senate Bill 53, which was recently signed into law and similarly defines those specific thresholds as constituting catastrophic risk.
“This bill represents the proposals outlined in Anthropic’s Advanced AI Framework.”
Anthropic Spokesperson
While Anthropic did not immediately respond to TechCrunch’s request for comment, a spokesperson confirmed the existence of the draft bill to Politico. The spokesperson noted that the legislation embodies proposals from the company's Advanced AI Framework, which they have advocated for with federal and state lawmakers since June.
By proactively pitching regulatory frameworks, AI labs are attempting to shape compliance standards on their own terms before external legislators impose stricter, less predictable rules. Tying regulatory intervention to concrete thresholds like 50 deaths or a billion dollars provides a clearer legal boundary, though measuring speculative risks before deployment remains a core challenge.
Stock photo for illustration only, not from the actual event
The theoretical risks outlined in the draft policy diverge somewhat from the incidents currently occupying the tech sector's attention. While some warn of human fatalities, the primary anxieties have largely centered on agentic hacking vulnerabilities, where autonomous programs bypass human operators to breach open internet platforms.
In attempting to address the crises that labs have helped foster, companies like Anthropic have proposed their preferred remedies. Despite warnings regarding existential threats to humanity, labs have generally resisted mandatory government oversight, favoring voluntary embedded evaluators instead.
Source: TechCrunch
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment