Anthropic Responsible Scaling Policy Version 3.0

• Authority: Private-sector companies
• Collections: Editors' Picks
• Most recent activity: Enacted
• Most recent activity date: 2026-02-24

Risk subdomain coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the risk subdomain are explicitly described.

• AI system security vulnerabilities
• Cyberattacks, weapon development or use
• Competitive dynamics
• Governance failure
• AI pursuing its own goals
• AI possessing dangerous capabilities
• Lack of transparency or interpretability

Minimal coverage — Brief mention but no specific mitigations, recommendations or governance measures relating to the risk subdomain are described.

• Disinformation, surveillance, and influence at scale
• Fraud, scams, and targeted manipulation
• Lack of capability or robustness

All other risk subdomains had no coverage in this document.

Risk Coverage Summary

The document has strong coverage of approximately 8-9 subdomains, with primary focus on malicious actors (4.1, 4.2, 4.3), AI system security (2.2), competitive dynamics (6.4), governance failure (6.5), and AI safety failures (7.1, 7.2, 7.3, 7.4). Coverage is heavily concentrated in security, misuse prevention, AI system safety, and governance domains related to catastrophic risks from advanced AI systems.

Sector coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the sector are explicitly described.

• Information
• Scientific Research & Development Services

All other sectors had no coverage in this document.

Sectors Summary

This is Anthropic's internal corporate governance policy, therefore it primarily governs the Information sector and Scientific Research and Development Services sector where Anthropic operates as an AI company. The policy does not regulate AI use in external sectors but rather governs Anthropic's own AI development, deployment, and safety practices that produce AI systems potentially used across various sectors.

Legislative Status

Other

Legislative Status Summary

This is an internal corporate policy document establishing a voluntary framework for AI safety governance. It is explicitly described as 'our voluntary framework for managing catastrophic risks' and represents company self-regulation rather than legally binding obligations or industry-wide standards.

Actors

Proposer

• Entity Types: AI_Developer
• Entities: Anthropic
• Summary: The document is Anthropic's internal Responsible Scaling Policy, developed and proposed by Anthropic itself as a voluntary framework for managing AI risks within the company.

Target

• Entity Types: AI_Developer
• Entities: Anthropic (primary target); Other frontier AI developers (secondary aspirational target for recommendations)
• Summary: The policy primarily applies to and regulates Anthropic's own AI development and deployment activities. It also includes recommendations for other frontier AI developers, but these are aspirational rather than binding.

Enforcer

• Entity Types: AI_Developer
• Entities: Responsible Scaling Officer (RSO); Board of Directors; Long-Term Benefit Trust (LTBT)
• Summary: Enforcement is internal to Anthropic through designated roles and governance structures. The RSO oversees implementation, the Board approves major decisions and policy changes, and the LTBT provides additional oversight.

Monitor

• Entity Types: AI_Developer, AI_Governance_Actor
• Entities: Responsible Scaling Officer; Board of Directors; Long-Term Benefit Trust; External reviewers (independent third-party organizations); Third-party compliance reviewers
• Summary: Monitoring involves both internal oversight by Anthropic's governance structures and external review by independent third parties who assess Risk Reports and procedural compliance with the policy.

AI Lifecycle Stages

Lifecycle stages covered by the documents, as defined in the OECD Framework for the Classification of AI Systems

• Plan and Design
• Build and Use Model
• Verify and Validate
• Deploy
• Operate and Monitor

AI Lifecycle Summary

The document comprehensively covers most AI lifecycle stages with primary emphasis on Plan and Design (through Frontier Safety Roadmap planning), Verify and Validate (extensive evaluation requirements), Deploy (deployment decisions and safeguards), and Operate and Monitor (Risk Reports, ongoing monitoring). Build and Use Model is addressed through capability thresholds and training decisions. Data collection is minimally covered.

Technical Scope

Terms relating to technical scope that are mentioned or implicitly referenced in the document

• AI Models
• Frontier AI
• AI Systems
• Compute Threshold
• Open Weight / Open Source

Technical Scope Summary

The document explicitly focuses on frontier AI models and highly capable models defined by specific capability thresholds. It extensively discusses AI models, AI systems, and model weights. It references compute scaling and training compute but does not specify exact FLOP thresholds. The document mentions open release of model weights. It does not explicitly use terms like GPAI, foundation models, generative AI, or predictive AI, though frontier AI systems are implicitly general-purpose.