🧭 Anthropic Transparency Hub — Safety Commitments, Model Cards & Policy Disclosures in One Place

Anthropic has launched a public Transparency Hub at anthropic.com/transparency — a consolidated resource that brings together safety commitments, model cards, usage policy summaries, third-party audit reports, and the company's responsible scaling policy documents under a single navigable interface. Previously these materials were scattered across the research blog, help centre, and GitHub, making them difficult to locate for journalists, policymakers, and enterprise procurement teams conducting due diligence.

What the hub contains

• Model cards — standardised capability and limitation summaries for every released Claude model
• Safety commitments tracker — a structured table showing Anthropic's public safety pledges and their current compliance status
• Third-party evaluations — summaries of external evaluator assessments, including the new audit pilot announced in the February Risk Report
• Policy changelog — versioned history of RSP updates, usage policy changes, and operator policy adjustments
• Incident log — a public-facing summary of significant safety incidents and how they were resolved (no PII or operator-identifying information)

Anthropic frames the hub as a response to growing calls from researchers and regulators for more structured AI accountability infrastructure, citing the EU AI Act's transparency requirements as one driver of the initiative.

🧭 RSP v3.0 — Policymaker Pre-Briefings Underway Ahead of Publication

Anthropic has confirmed that pre-publication briefings on Responsible Scaling Policy v3.0 are now underway with a select group of government officials and safety researchers. The company has chosen to brief key stakeholders before the full public release to allow policymakers to incorporate the updated framework into ongoing AI governance work. This is the first time Anthropic has conducted pre-briefings at this scale ahead of an RSP revision.

While the full text of RSP v3.0 has not been released, Anthropic has confirmed that it incorporates the updated ASL-3 trigger criteria from the February Risk Report, expands the scope of mandatory third-party evaluation, and introduces a new Deployment Review Board — an internal governance body with authority to delay model releases pending additional safety review. Publication of the full document is expected within the next two weeks.