Recorded Future counts 103 AI-enabled attacks in a year and finds fully automated ones still rare

Recorded Future Insikt Group · Research report

Read the source (opens in a new tab)
What happened
Recorded Future's Insikt Group published an analysis on November 11, 2026 of 103 in-the-wild incidents between July 31, 2025 and July 31, 2026 in which AI was the attack vector, tool or target.
Key points
  • Most incidents used AI to speed up familiar attacks such as supply-chain compromise, phishing and extortion; AI developer tools are "a prime target" because they hold credentials and code.
  • Of the 103, 32 used AI in malware or to run operations, and most state-sponsored activity sits at Level 2 to 3 of its AI Malware Maturity Model.
  • The exception it names is a campaign against Taiwanese government entities that used up to eight AI agents with limited human intervention.
  • Insikt links the rarity of full automation to token costs and the need for human oversight, and expects autonomous and orchestrated campaigns to become more frequent.
Evidence
A security vendor's research report counting incidents documented in its own intelligence graph, excluding proofs of concept and security research. The campaign rests on Dream Research's report, which it links; that report describes government entities in Asia and does not name Taiwan.
Uncertain
The count covers only incidents searchable in Recorded Future's own intelligence graph.

See it in the latest record