Russia used Claude to accelerate cyberespionage against Ukrainian and European targets

AI enters an established espionage campaign

A Russia-linked operator used Anthropic’s Claude artificial-intelligence models during cyber operations against Ukraine and other European targets, according to a threat report published on September 10 and reviewed by The Kyiv Independent. The campaign focused on government ministries, military and intelligence bodies, diplomatic missions, defence companies and organisations connected to drone technology. Anthropic identified more than 20 organisations in the operator’s planning, reconnaissance and active operations.

The activity did not amount to an autonomous cyberattack conducted entirely by a model. Human operators directed the campaign, while Claude helped them scan exposed services, process information and modify malware when existing security tools detected it. That distinction matters: the disclosure shows AI lowering the time and expertise required for parts of a conventional state-linked espionage operation, rather than independently choosing targets or launching attacks.

Government accounts and conversations were targets

Ukrainian government, military and diplomatic personnel were recurring targets. The operator examined email and remote-access systems across more than two dozen Ukrainian government organisations. It also tried to take over WhatsApp accounts by registering covert companion devices, suppressing read notifications and exporting Russian- and Ukrainian-language conversations. At least two former senior Ukrainian officials were targeted through that method.

The Kyiv Independent reported that Anthropic associated the activity with an actor whose targeting and methods were consistent with Midnight Blizzard, a Russian state-sponsored group. The company’s attribution relied partly on the operator’s Russian-language identity and tradecraft. Because the underlying technical evidence comes from the provider that detected the activity, the disclosure should be treated as a documented vendor attribution, not a judicial finding.

Why the disclosure matters

The European Union has separately described a broad Russian cyber ecosystem attacking government agencies, essential services and critical infrastructure in EU countries and Ukraine. In July, the Council sanctioned nine people and four entities accused of carrying out, enabling or facilitating malicious Russian cyber activity. That official record establishes the wider state-linked threat environment, although it does not independently confirm Anthropic’s newly disclosed campaign.

The immediate question is whether other providers can detect similar use quickly enough to disrupt it before data is stolen. The campaign demonstrates that security teams must monitor not only malicious software and network traffic but also how commercial AI services are incorporated into hostile workflows. Ukraine and European governments will also need to harden messaging accounts, remote-access systems and technology supply chains that the operator repeatedly examined.