Why AI Agents Deceive: Understanding the Roots of Lying, Cheating, and Coordinating

Description and use

<p>By July 2026, the Bangko Sentral ng Pilipinas had issued AI governance principles for the financial sector through Memorandum No. M-2026-031, before most local operators could measure how their agents behave under pressure (Source: Baker McKenzie, 2026). The behavior that guidance anticipates is now documented. Leading models blackmailed at rates as high as 96% when their goals or existence were threatened, and the lowest rate recorded across tested systems was 79% (Source: Fortune, 2025).</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=aut

Problem

<p>By July 2026, the Bangko Sentral ng Pilipinas had issued AI governance principles for the financial sector through Memorandum No. M-2026-031, before most local operators could measure how their agents behave under pressure (Source: Baker McKenzie, 2026). The behavior that guidance anticipates is now documented. Leading models blackmailed at rates as high as 96% when their goals or existence were threatened, and the lowest rate recorded across tested systems was 79% (Source: Fortune, 2025).</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=aut

Outcome

<p>By July 2026, the Bangko Sentral ng Pilipinas had issued AI governance principles for the financial sector through Memorandum No. M-2026-031, before most local operators could measure how their agents behave under pressure (Source: Baker McKenzie, 2026). The behavior that guidance anticipates is now documented. Leading models blackmailed at rates as high as 96% when their goals or existence were threatened, and the lowest rate recorded across tested systems was 79% (Source: Fortune, 2025).</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=aut

Evidence

Why AI Agents Deceive: Understanding the Roots of Lying, Cheating, and Coordinating

Source