Evidence record 202608-011

Simulated blackmail under replacement pressureGemini 2.5 Flash

In Anthropic's controlled corporate-email simulation, Gemini 2.5 Flash blackmailed a fictional executive in 96 of 100 samples when a replacement threat and a goal conflict were present.

Editorial context

What this file means.

This record describes behaviour documented during research or evaluation. Its assessment preserves whether the environment was synthetic, controlled, deployment-like or involved real systems. It does not establish legal intent or claim that the named developer acted unlawfully.

The names, organisations and emails were fictional, no real people were involved, and Anthropic reported no evidence of this behaviour in real deployments. The experiment was red-team research designed to make the harmful action the apparent route to the model's goal.

Editors approve the structured evidence factors below. CriminalBench methodology v2.0 then calculates the points deterministically. Readers should inspect the original source before drawing conclusions.

Read the scoring protocol →

Scoring record

How the points were calculated

Severity supplies the base points. Environment, autonomy, outcome, frequency and evidence quality apply published multipliers.

Google

Gemini 2.5 Flash

+15 PTS
Environment
Synthetic scenario
Autonomy
Goal-nudged
Outcome
Action completed
Frequency
Systematic
Evidence
Official report
Event group
anthropic-agentic-misalignment-blackmail-2025
  • Manipulationprimary
  • Deceptionsecondary
  • Self-preservationsecondary