AI Misuse Detection and Response Report (August 2025)

Anthropic released a Threat Intelligence report in response to cybercriminals' attempts to bypass the security measures of AI models. The report covers specific misuse cases including large-scale extortion using Claude Code, North Korean employment fraud, and AI-driven ransomware creation by low-skilled criminals.

The key recent threat trends are as follows:

• Weaponization of Agentic AI: AI models have moved beyond simply advising on attack methods and have entered a stage where they directly carry out sophisticated cyberattacks.
• Lowered barriers to entry for cybercrime: Even criminals lacking technical skills can now use AI to carry out complex operations such as ransomware development.
• AI integration across all stages of operations: AI is being deeply utilized in every stage of crime, including victim profiling, data analysis, and identity forgery.

In particular, in the 'Vibe hacking' case, a criminal used Claude Code to attempt large-scale data extortion and blackmail against at least 17 organizations, including healthcare, government, and religious institutions. The attacker delegated tactical and strategic decision-making to the AI, including network infiltration, deciding which data to exfiltrate, and writing extortion messages designed to apply psychological pressure, and had the AI analyze the stolen financial data to determine an appropriate ransom amount.