Anthropic has submitted a draft federal AI safety bill to US lawmakers, proposing a threshold of 10²⁵ FLOPs to define covered AI models. By June 2025, at least 33 models from 12 developers had already crossed this threshold, a number expected to keep growing. Critics argue this threshold reflects an earlier era of AI and could capture far more models than intended, especially compared to California’s SB 53, which sets its trigger ten times higher at 10²⁶ FLOPs. To qualify as a covered developer, a company must also meet financial thresholds of over $500 million in annual AI-derived revenue or over $1 billion in annual AI R&D spending. The bill limits government intervention to catastrophic scenarios defined as damages exceeding $1 billion or loss of more than 50 lives.
Source: Read the original article
