00:00Anthropic says it disrupted several malicious uses of its Claude AI models.
00:04The company says the activity involved both cybercrime and state-linked operations.
00:09Anthropic identified attacks connected to seven China-based AI labs.
00:13Those included Alibaba, Moonshot, DeepSeek, and Xiaomi.
00:17Alibaba-linked operators allegedly generated more than 151 million exchanges with Claude.
00:23The activity was recorded between May and July 2026.
00:26It peaked at nearly 3 million exchanges in a single day.
00:30Anthropic says some groups used Claude responses to train competing AI systems.
00:35The company also identified a suspected Russia-linked cyberespionage campaign.
00:40The group allegedly used AI for phishing and hotel Wi-Fi hijacking.
00:44It also targeted WhatsApp accounts connected to Ukrainian officials and organizations.
00:49Anthropic says AI was used throughout much of the attack process.
00:52The report also identified misuse involving software for weapons programs.
00:57The findings show how increasingly capable AI models can create new cybersecurity risks.
01:02Disclosure.
01:03This video contains stock footage and content created or enhanced using AI-assisted tools.
Comments