Anthropic, by updating its policies, has declared access to Claude AI by individuals and companies based in Iran or under their control to be unauthorized.
Details of the access restrictions to Claude in foreign countries can be seen in the new version of Anthropic's policies, according to which service coverage for these countries will change. Iran is also one of the countries present on this list. Based on these new rules, the use of Claude is not permitted for individuals present in Iran, companies registered or headquartered in the country, and entities that are majority-owned or controlled by individuals and entities based in Iran.
Of course, these restrictions will be implemented in a few days. The creator of Claude has announced that the new policy it has set for the use of its services will be implemented from November 12, 2026. According to Anthropic, most of the changes have been made to clarify existing rules and explain how they apply to new Claude capabilities.
Registering a company outside of Iran is not enough either
The section related to unsupported regions includes three groups: individuals who are present in these regions when using these services; companies and entities that are registered or have headquarters there; and the third group, entities that are directly or indirectly majority-owned or controlled by individuals and entities in these regions. The restrictions on the latter two groups also apply to their users and employees even when they are present in supported countries.
Therefore, based on the policy text, registering a company in another country is not necessarily sufficient to access the services; the structure of its ownership and control also matters. Of course, being Iranian is not the sole criterion for prohibition; the criteria relate to the location of presence, place of registration or headquarters, and the structure of ownership and control. This policy covers Anthropic's products and services, including applications, APIs, and models offered through cloud platforms and authorized vendors.
Anthropic has introduced this change as a clarification of how the previous policy is implemented. In September 2025, the company had also announced and implemented restrictions for companies controlled by entities based in unsupported regions; these restrictions applied to companies where more than 50 percent of ownership, directly or indirectly, was held by companies with headquarters in the targeted regions.
Combating Influence Operations and Fake Account Networks
The new update implemented by the company, in addition to geographical restrictions, also integrates rules regarding deceptive operations. What does this mean? Anthropic says that in the past year, it witnessed instances of state media usage where government propaganda offices and commercial companies used Claude to manage networks of fake accounts and fabricated news websites. Now, the new regulations explain the concealment of message originators' identities, artificial amplification of content, and the creation of tools for executing influence operations in a separate section.
The company has also updated its policies regarding the ban on using Claude for weapons development. Under this rule, software for weapon guidance and control, components required for their operation, and the arming of drones or other autonomous vehicles are also included in the prohibition list.
Restrictions Aimed at Individual Surveillance and Sensitive Decision-Making
In the surveillance and law enforcement section, Anthropic has emphasized that tracking individuals without consent, whether in real-time or through historical data analysis, is prohibited. Using Claude to make decisions or provide recommendations on who should be investigated, detained, or charged is also no longer permitted.
Limitations have also been considered for a series of high-risk applications. In areas such as healthcare, law, and finance, human oversight has been made mandatory. According to the new rules, this human overseer must be a qualified professional who has the authority to review and modify Claude's recommendations.
Additionally, the individual affected by the decision must be informed of the use of artificial intelligence. In applications connected to equipment with autonomous physical movement and the potential for causing harm, a qualified operator must be able to observe the device's operation and stop it if necessary.
Another change mentioned in Anthropic's policies is severe, repetitive, and purposeless abusive behavior toward the company's models. These behaviors will be prohibited under the new policies starting November 12.
According to the company, this rule does not cover typical user dissatisfaction, disagreement with responses, dark themes in creative works, or testing and research on models, and Claude terminating such conversations will be the primary enforcement mechanism.
Comments