Australian Enterprises At Risk as Anthropic Finds Hackers In Claude Code

In September 2025 a Chinese state-linked group leveraged Anthropic's Claude Code to run highly automated intrusion attempts against roughly 30 organizations, with the agent performing 80–90% of actions; the report warns Australian organisations that widespread agent adoption and weak governance (inventory, identity, and revocation controls) create a critical attack surface and recommends inventorying agents, routing access through policy enforcement, treating agents as non-human identities with least privilege and expiry, and building tested shutdown procedures.