AI Boosts Cybercrime, Poses New Challenges for Defenders

Key Points

• Generative AI can produce flawless phishing emails, raising click‑through rates from ~12% to >50% in tests.
• Voice‑cloning and deep‑fake attacks surged 442% and 680% year‑over‑year between 2023‑2024.
• Prompt‑injection attacks embed hidden commands in content, tricking AI assistants into leaking data.
• Meta’s Instagram support bot allowed attackers to hijack accounts lacking multi‑factor authentication.
• Microsoft’s Project Perception deploys AI agents for red‑team, blue‑team and remediation tasks.
• Defenders must protect every vulnerability, while attackers need only find one to succeed.
• AI‑generated code shows higher rates of high‑risk bugs than manually written code, even after review.