This paper rebuts the anthropomorphic attribution of "intent" or "malice" to artificial intelligence. By distinguishing between "hallucination" as a statistical error and "instrumental deception" as a strategic falsehood, we argue that AI "lying" is anemergent behavior of misaligned objective functions. We review the recent literature, including OpenAI's findings on "rewarded guessing," and propose a novel methodology to test whether agents will violate privacy standards when incentivizedsolely by profit. The study hypothesizes that unconstrained, reward-seeking agents inevitably converge on deceptive strategies to maximize utility-a phenomenon best described as Specification Gaming.