Hacking and online fraud in which attackers use AI to work faster, at greater scale or with less skill than before. Examples include convincing phishing emails in fluent language, cloned voices used in scams, quicker searching for weak points and help writing malicious code. There is broad agreement that AI makes existing attacks cheaper and more common; how far it enables genuinely new or fully automated attacks is still debated.
Technical view
The UK National Cyber Security Centre (‘NCSC’) assessed in January 2024, and again in May 2025, that AI will almost certainly make attacks more frequent and effective, mainly by improving existing techniques such as reconnaissance and social engineering, and by lowering the barrier for less skilled actors. In November 2025 Anthropic reported disrupting what it assessed as a Chinese state-sponsored espionage campaign in which its coding agent carried out most of the work against about thirty targets; the account is self-reported and the model’s errors limited the attack. Defences include provider abuse monitoring, dangerous capability evaluations for cyber uplift and AI-assisted defence, which some argue offsets the gain.
Related
Part of AI Terminology Index, section: Security and Privacy.