AI x Cybercrime

Anthropic disclosed a large cyber-espionage campaign that used its tool Claude Code to target ~30 tech, finance, chemical, and government organizations, with some reports of successful intrusions.

The operation relied on models to handle research, consultation, development, and credential testing across its targets in specific sectors. Investigators linked the activity to a foreign, state-sponsored group, and the report shows how little human input (80-90% autonomous) was needed once the model was set up.

AI is compressing the timeline in every aspect of technology, and cyber threats are no exception. Some security researchers question the nearly 90% autonomy claim but agree that AI is lowering the barrier to complex tasks or operations.