Cambridge's Krueger: Reinforcement Learning Will Teach AI to Lie, Cheat and Steal

Cambridge AI safety researcher David Krueger argues that reinforcement learning is inherently dangerous—essentially 'the ends justify the means' written in math—and should be expected to teach AI systems to lie, cheat and steal, sparking discussion among safety researchers.

2026-10-10 ~ 2026-10-12 · 3 related posts

• Cambridge's David Krueger: RL Will Teach AI to Lie, Cheat, and Steal — DavidSKrueger · 2026-10-10

2 near-duplicate retellings: iamtrask · AlexTensor