Anthropic Updates Responsible Scaling Policy to Version 3.2
On April 29, 2026, Anthropic released Version 3.2 of its Responsible Scaling Policy. The update authorizes the company’s Long-Term Benefit Trust (LTBT) to request external reviews of Risk Reports, approve the selection of external reviewers, and requires regular briefings for the LTBT.
TLDR
Anthropic updated its Responsible Scaling Policy (RSP) to Version 3.2, effective April 29, 2026. The changes strengthen oversight by the Long-Term Benefit Trust (LTBT), giving it authority to request external reviews of Risk Reports, approve external reviewers, and receive regular briefings on the policy’s implementation.
Key Updates in Version 3.2
The RSP is Anthropic’s voluntary framework for managing catastrophic risks from advanced AI systems. Version 3.2 builds on prior iterations by enhancing independent oversight mechanisms:
- The LTBT can now request external review of Risk Reports.
- The LTBT has the power to approve Anthropic’s selection of external reviewers.
- The policy formalizes a requirement for the LTBT to receive regular briefings.
These provisions aim to increase transparency and accountability in how Anthropic evaluates and mitigates risks as models advance.
The full RSP document and redline comparing versions are available on Anthropic’s site. Earlier versions (3.1 effective April 2, 3.0 effective February 24) introduced Frontier Safety Roadmaps and Risk Reports.
Why this story matters
As frontier AI capabilities continue to advance, voluntary governance frameworks like Anthropic’s RSP serve as important case studies for industry self-regulation. The Version 3.2 updates specifically bolster the role of the LTBT—an independent body—in overseeing risk assessments. This reflects ongoing efforts to balance rapid development with structured external scrutiny, particularly around high-stakes areas such as AI R&D automation and potential misuse risks. The changes provide concrete details on how one leading lab is evolving its risk management processes.
Sources
- Anthropic: “Anthropic’s Responsible Scaling Policy” (Version 3.2 effective April 29, 2026). https://www.anthropic.com/responsible-scaling-policy
- RSP document and redline for Version 3.2.
Featured Image Alt Text
Anthropic logo with scaling policy document and oversight icons representing the Version 3.2 Responsible Scaling Policy update announced April 29, 2026
Tags
Anthropic, Responsible Scaling Policy, RSP, AI Safety, Governance, LTBT