Anthropic is pushing Congress to mandate independent safety tests for powerful AI models
Quartz
—
The company wants independent evaluations required for models trained above a compute threshold, with government authority to block dangerous deployments