Anthropic activates AI Safety Level 3 protections for Claude Opus 4
Action Required
Failure to migrate to a supported model or adhere to ASL-3 restrictions could lead to misuse of Claude Opus 4 for CBRN-related activities.
AI Impact Summary
Anthropic is proactively activating AI Safety Level 3 (ASL-3) protections for Claude Opus 4 as a precautionary measure due to anticipated advancements in CBRN-related knowledge and capabilities. This represents a significant shift from the baseline AI Safety Level 2 (ASL-2) and demonstrates a proactive approach to mitigating potential misuse, particularly concerning weapons development. While the model's capabilities haven't definitively surpassed the ASL-3 threshold, the company is implementing enhanced deployment and security controls to address evolving threats and ensure responsible AI development.
Affected Systems
- Date
- Date not specified
- Change type
- capability
- Severity
- high