Anthropic Reveals 7 Risk Areas from Claude Abuse...AI Key Theft and Unreleased Model Access Attempts

Anthropic has revealed seven risk areas associated with the malicious use of Claude, including state-backed espionage, data theft, and AI supply chain attacks. In its latest threat intelligence report, Anthropic stated it identified and blocked malicious activities between December 2025 and August 2026.
The report noted that attackers suspected to be linked to Russia targeted over 20 government, defense, and diplomatic organizations in Ukraine and Europe using Claude. Additionally, attackers associated with Shaihunters reportedly stole AI API keys in enterprise environments to target other organizations, though Anthropic's systems were not compromised during these incidents.
Anthropic also reported that one attacker attempted to target approximately 30 AI companies over four days, seeking to access unreleased Claude models using stolen keys from client accounts. All attempts failed. The report further categorized surveillance, fraud and scams, biological misuse, conventional weapons development, and illegal model distillation as key risk areas.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 클로드 악용 7개 위험 영역 공개…AI 키 탈취·미공개 모델 접근 시도