Anthropic requires a human in the loop

Anthropic published an updated usage policy on October 8 that takes effect on November 12, 2026. “When our products are used in ways that can affect someone’s health, legal rights, finances, livelihood, or access to essential services, we require there to be a qualified ‘human in the loop,’” the company wrote.

The policy defines that person as someone with the authority to review and, if necessary, change Claude’s recommendations. It also says the people affected must be told that AI was used in the decision.

The terms bind the vendor that builds on Claude, and through it the bank that uses the product. A credit, collections or account-decision tool that runs on Claude now carries a contractual human-review requirement from its model provider.

What this means for your bank

Ask vendors whose credit or account-decision tools run on Claude how they will meet the policy by November 12.
Check that your adverse-action and disclosure practices tell customers when AI was used in a decision.

Sources

2026 usage policy update, October 8, 2026 (Anthropic)anthropic.com ↗

This is for information only and is not legal advice. Verify the policy terms against the linked page before acting.

Keep reading

Anthropic heads for a November IPO at up to $2 trillion

Bank examiners start asking about AI agents and Chinese models

Browse the full Regulatory Radar