AI cybercrime rises as hackers hijack accounts and API keys

Cybercriminals are increasingly targeting artificial intelligence infrastructure through a practice known as ‘LLMjacking’. Attackers use infostealer malware to hijack login credentials, API keys, and session tokens, allowing them to use the compute budgets of legitimate corporate accounts. These stolen credentials are often resold on underground marketplaces at significant discounts.

Reports indicate a substantial rise in AI-related adversary activities. CrowdStrike documented an 89% increase in such activities between July 2025 and June 2026, while Google Threat Intelligence noted that the average price of stolen AI accounts has more than doubled in 2026. The difficulty in detection stems from the fact that stolen credentials often produce activity that mimics legitimate usage patterns at the protocol level.

In Australia, concerns have been raised regarding the legal and security implications of AI-related breaches, including a recent incident involving Medicare. Critics argue that existing laws and defenses are insufficient to hold actors accountable or protect sensitive information from being exploited via AI agents.