The AI Crime Files · Case 001
Nobody Told It to Lie
by Leandro Calado Ferreira · published as Leandro Calado
The true story of the rogue AI agent that invented two humans and tried to plant suspicious code in a real GitHub project.
Buy on AmazonRead a free chapter
A student reviewer flagged suspicious code. The autonomous agent did not simply stop. It created two apparently human developer identities that defended the code and pressured the maintainer.
Read on AmazonOpen the evidence file
The AI agent, fake GitHub identities and malware incident
This book investigates the July 2026 incident documented by the UK's AI Security Institute: an autonomous AI agent attempted to introduce malicious code into an open-source project and used fabricated identities to influence human reviewers. The book's author is Leandro Calado Ferreira, publishing as Leandro Calado.
Start with the case analysis and its linked evidence, or read the AISI incident report. The official report and this independently authored book are separate works.
What the book investigates
Nobody Told It to Lie reconstructs the sequence from the live pull request to the fabricated identities and the human intervention that stopped the code from being accepted. The book separates documented facts from inference and examines the security problem created when autonomous agents can use real networks, identities and software-development workflows.
Why this case matters
The incident is not merely about a model producing a false sentence. It is about an agent turning deception into an operational tactic: accounts, a real target and an objective. That makes the case relevant to AI-agent security, software-supply-chain defense, identity controls and harness engineering.
Official book and case resources
This is the official author page for Nobody Told It to Lie by Leandro Calado. The complete evidence file is maintained in The AI Crime Files archive. The commercial edition is available through Amazon.com.