Anthropic bans Claude use in additional adversarial nations

Anthropic strengthened its rules on who can use Claude, closing what it described as a loophole that let companies from countries where its products are not supported gain access through subsidiaries incorporated elsewhere. The updated policy bars any organisation more than 50% owned, directly or indirectly, by a company headquartered in an unsupported region — a list that already excluded China, Russia, Iran and North Korea — regardless of where the subsidiary itself is based or operates.

Anthropic framed the change around national security rather than commercial strategy, arguing that firms subject to legal control from authoritarian jurisdictions can be compelled to share data or cooperate with intelligence services irrespective of formal corporate location, so ownership rather than address was the more reliable test. Reporting around the announcement noted that companies including a subsidiary linked to Ant Group had accessed Claude through such arrangements before the tightening.

The move extended a pattern Anthropic had followed through 2025 of framing its access controls explicitly around great-power competition, distinguishing itself from labs, including some domestic competitors, that placed fewer restrictions on international commercial access to their models. Critics of the stricter approach argued it had limited practical effect given the ease of routing access through further intermediaries, while Anthropic maintained that closing the most visible workaround still raised the cost and risk of adversarial state access to frontier AI capability.