Back
Automatically translated version. May contain inaccuracies compared to the original.
Russian hackers attempted to use the artificial intelligence Claude from Anthropic to conduct cyberattacks against the Ukrainian government, military, and diplomatic structures.
Anthropic released Threat Intelligence, a report documenting abuses of its AI technologies over the past eight months. According to the developers, malicious actors are increasingly using AI to organize cyberattacks, leaving humans with only supervisory roles. How hackers attacked Ukraine A state-backed hacking group used Claude models at nearly every stage of its operations. They conducted phishing campaigns, intercepted Wi-Fi data in hotels, and gained access to WhatsApp accounts. All of these attacks targeted Ukrainian government, military, and diplomatic officials. By leveraging AI capabilities, the perpetrators built a system that automatically detected the subtraction of malware by antivirus and rewrote it until it became "invisible" to security tools. Analysts note that the group’s tactics align with the operations of the Russian group Midnight Blizzard, which is linked to Russia’s Foreign Intelligence Service. Not only hacking, but also weaponry and data leakage In addition to cyberattacks against Ukraine, the company identified other threats: Weapon development - Users from Russia, China, and Yemen attempted to use Claude to create software for firearms, missiles, combat drones, and guidance systems. Bioweapons development attempts - Experts blocked several accounts where researchers tried to plan experiments to adapt mammals to avian flu. Attacks on Chinese competitors - Seven Chinese laboratories (including Alibaba, Moonshot, and DeepSeek) carried out mass attacks to intercept Claude responses and use them to train their own AI models. All identified attacker accounts were blocked, and the obtained data used to strengthen Anthropic’s product protections.
Russian hackers attacking Ukraine By the way, this is not the first time Russian hackers have exploited technological vulnerabilities to attack Ukrainian structures. Recently, cybercriminals from the Russian Federation found a way to blind the AI protection during attacks on Ukrainian organizations. The technique, called GuardBreaker, caused language models to fail to verify files, enabling Russian hackers to "break" AI defenses and leave malware undetected during attacks on power and transportation sectors. In addition, Google Threat Intelligence Group researchers uncovered a spying .NET backdoor STOCKSTAY. Using this program, Russian hackers attacked government agencies and military departments of Ukraine, as well as European diplomatic institutions, masking the virus as legitimate utilities.
Document: PDF proof of the original version of the news item "Для кібератак на Україну росіяни намагалися використати Claude, - Anthropic". It records the publication content at the moment of the first scan, the preservation date and the source: RBC-Ukraine.
