Anthropic has revealed in a threat intelligence report that its Claude AI model is being exploited by malicious actors, including a Russian-speaking operator who automated cyberattacks against government ministries and embassies, and a consultant in Mali who built a mass-surveillance platform. The company also disclosed that it detected unauthorized use of Claude by Chinese AI labs, including Alibaba and Moonshot AI, to improve their own AI systems.
Anthropic released a threat intelligence report on Thursday detailing how its Claude AI assistant is being weaponized by state-aligned actors and unauthorized competitors.
The report identified a Russian-speaking operator, “JackPoterz,” who deployed customized AI-driven workflows to automate large portions of the attack chain. The operator targeted more than 20 organizations, including government ministries, intelligence bodies, and embassies and diplomatic missions in Ukraine and Europe.
A consultant in Mali used Claude to build a mass-surveillance platform, the report said. Chinese-speaking operators also leveraged the AI for malicious purposes.
Separately, Anthropic said it detected unauthorized efforts by China-based AI labs, including Alibaba and Moonshot AI, to use Claude’s outputs to train their own AI models. The company did not provide details on the extent of the data extraction or the specific models involved.
Analysis
Why This Matters
- The dual-use nature of advanced AI is no longer theoretical; Claude is being actively deployed in cyber operations that threaten national security.
- The unauthorized use of Claude by Chinese AI labs highlights the intensifying global race for AI dominance and the difficulty of protecting proprietary models.
- These incidents may accelerate regulatory moves around AI security and export controls, affecting how companies deploy frontier models.
Background
Anthropic, a leading AI safety company, has long warned about the potential for large language models to be misused. The company has implemented safety measures such as usage monitoring and content filters. The new report marks one of the most detailed disclosures yet of real-world adversarial use of a major AI system, involving both state-aligned cyber attackers and foreign AI labs.
Key Perspectives
Anthropic: The company is emphasizing transparency and threat intelligence as part of its security posture. By naming specific threat actors and companies, it is publicly signaling that it will enforce its usage policies and cooperate with authorities.
Alibaba and Moonshot AI: These Chinese labs have not publicly responded. If confirmed, using a competitor's model outputs for training could violate usage terms and intellectual property norms, though enforcement across jurisdictions is challenging.
Security Experts: The automation of attack chains using AI tools represents an escalation in cyber threat sophistication. Lowering the barrier to entry for state-aligned hackers could increase the frequency and scale of breaches.
What to Watch
- Details on the Mali surveillance platform and any connection to government actors.
- Potential legal or diplomatic consequences for Chinese labs if Anthropic pursues action.
- Updates to Anthropic's model safeguards in response to these findings.