Anthropic details AI usage rules; new version takes effect in November
The update brings together examples for deceptive campaigns and high-risk uses. Among the controls spelled out are qualified human review and notice to affected people, requirements that, according to the company, already existed. The new version of Claude's policy takes effect on November 12; it is not a new Brazilian law.
A version to make the limits clearer
The Usage Policy update follows the expansion of the tasks Claude can carry out with greater autonomy. According to Anthropic, most of the changes clarify existing rules. Among them is a specific section on deceptive campaigns or artificial activity, including hiding who is behind a message, amplifying content through fake accounts or building infrastructure for influence operations.
The company also reorganized the section on elections, detailed further the prohibitions related to weapons and surveillance and started describing controls for models connected to equipment that performs physical actions. The revised version takes effect on November 12, 2026.
What the policy already required for high-risk uses
When Anthropic products are used in situations that can affect health, legal rights, finances, livelihoods or access to essential services, the policy requires a qualified person with the authority to review and change the model's recommendations. The affected person must also be told that AI was used. Anthropic says these requirements were already in force; the update explains them more precisely and includes new examples.
This distinction matters: clearer wording may require reviewing the internal routines of whoever uses the product, but it is not, by itself, a new law for Brazilian companies. The policy describes the terms of use of Anthropic's products; the organization still needs to check the laws, contracts, data and responsibilities that apply to its case.
How to prepare the team
Before a pilot with Claude, the company can list the intended uses, identify whether a recommendation affects people and define who will have the power to review, correct or stop the decision. Training needs to include real examples from the role, criteria for escalating a case and a record of human reviews. Updating a document without practicing these scenarios tends to leave operational doubts.
Anthropic's post is a source from the vendor itself, not an independent legal analysis. The updated text and the applicable terms should be checked before deployment.
Sources
Anthropic, 8 Oct 2026. https://www.anthropic.com/news/2026-usage-policy-update
Previously published on dg5i.com.br.