Responsible Scaling

Oct 14, 2026

• Anthropic

  Reflections on our Responsible Scaling Policy

  Anthropic reflects on implementing its first Responsible Scaling Policy, which aims to turn high-level safety concepts into practical guidelines for addressing catastrophic failures from frontier AI models. The policy has clarified organizational priorities, framed discussions around timelines and headcount, and surfaced dependencies that might otherwise have gone unaddressed. The five commitments include identifying and publishing red line capabilities, constructing frontier risk evaluations as empirical tests, responding with scaled safety and security standards (ASL-3/4), iteratively extending the policy with higher standards, and implementing assurance mechanisms with board oversight and external review. Currently, teams test models pre-deployment in cybersecurity, CBRN, and model autonomy when compute reaches four times the most recent tested model, and retest every three months to account for finetuning improvements. Anthropic draws insights from nuclear security, biosecurity, systems safety, autonomous vehicles, aerospace, and cybersecurity, finding each domain offers valuable but imperfect analogies for frontier model risk management.

• Anthropic

  Introducing Anthropic's Transparency Hub

  Anthropic has launched the Transparency Hub, a detailed resource outlining measures to ensure systems are safe, beneficial, and trustworthy. The hub publishes a first periodic report covering banned accounts, account appeals, appeal overturns, reports to the National Center for Missing and Exploited Children, and government request data. It also provides information on methodologies for model evaluation, abuse detection, internal governance, societal impact assessment, AI safety research, and privacy safeguards. Anthropic designed this hub as a unified framework to address diverse transparency requirements, giving stakeholders a clear view into model development and deployment. The company commits to continuously expanding reporting to reflect evolving best practices as AI capabilities advance, and it welcomes feedback.

• Anthropic

  Announcing our updated Responsible Scaling Policy

  Anthropic published a significant update to its Responsible Scaling Policy (RSP), a framework for managing catastrophic risks from frontier AI. The update introduces two key capability thresholds—autonomous AI research and development and CBRN weapons—that would require upgraded safeguards from ASL-2 to ASL-3 or higher. It refines model evaluation and safeguard assessment methodologies by adopting safety case methods from high-reliability industries and adding internal governance and external expert feedback. The policy reaffirms models will not be trained or deployed without adequate safeguards and incorporates lessons from minor procedural issues like late evaluations that posed minimal risk. Co-Founder Jared Kaplan becomes Responsible Scaling Officer, with the company hiring for RSP teams including Frontier Red Team and Trust & Safety.