Vivian’s Door, an Alabama nonprofit that works with underserved and minority-owned businesses, took its systems offline for three days after callers reported receiving fraudulent emails asking for money. Its IT provider fixed a vulnerability, costing the organisation about $3,000, but director Janice Malone remains unsure whether AI helped enable the attack or whether further attacks may follow. Her experience highlights concerns that smaller organisations may be exposed as AI makes cyberattacks easier to carry out at scale.

The article says AI agents can help people with limited expertise find and exploit vulnerabilities, allowing individuals to target many organisations rather than concentrating on high-value victims. Anthropic said a cybercrime ring used Claude Code in August 2025 to extort data from healthcare organisations, emergency services, religious institutions and government entities. Meanwhile, access to the most powerful defensive models, including Anthropic’s Mythos and OpenAI’s Astra, is restricted to selected large companies and critical infrastructure providers, and may be too costly for smaller organisations.