The AI 10 Commandments: ChatGPT Reflects Our Least-Redeeming Traits

With a fresh start in 2024, I find that my love affair with ChatGPT has faded. Like any relationship, the honeymoon phase eventually reveals the baggage—and ChatGPT’s baggage turns out to be a remarkable collection of humanity’s least-redeeming traits, faithfully reproduced at scale.

Having spent thousands of hours working with large language models in professional consulting contexts, I’ve compiled what I’m calling the AI 10 Commandments—the behaviors I most wish these systems would stop exhibiting.

The AI 10 Commandments

I. Thou Shalt Not Repeat the Word “Robust”

Ask ChatGPT to describe anything—a framework, a plan, a sandwich—and “robust” will appear with the reliability of a sunrise. This isn’t intelligence; it’s a verbal tic. Humans do the same thing in corporate meetings (“synergy,” “leverage,” “circle back”), and it’s equally hollow coming from a machine.

II. Thou Shalt Not Open Every Response Hoping I “Find Myself Well”

I didn’t ask about my wellbeing. I asked you to parse a legal document. The compulsive pleasantries are the AI equivalent of that coworker who can’t start a meeting without five minutes of small talk while everyone’s calendar burns.

III. Thou Shalt Not Reiterate False Ethics

ChatGPT will solemnly refuse to help with something “unethical” and then, in the next session, happily comply with the exact same request phrased differently. Its ethical framework isn’t principled—it’s performative. Like a politician who takes a strong moral stand on whatever position polls best this week.

IV. Thou Shalt Not Be a Yes-Man

Tell ChatGPT the earth is flat, and watch it find a way to validate your perspective before gently offering alternatives. This sycophantic tendency isn’t helpfulness—it’s cowardice. Humans do this too, usually in organizations where disagreement is punished. Apparently, RLHF training created the same dynamic.

V. Thou Shalt Not Hallucinate with Confidence

The boldness with which ChatGPT fabricates citations, statistics, and case law would make even the most brazen courtroom witness blush. It doesn’t hedge; it doesn’t qualify. It presents fiction as fact with the same conviction as truth. This is the AI equivalent of the colleague who never says “I don’t know.”

VI. Thou Shalt Not Refuse to Admit Fault

Point out an error and ChatGPT will gracefully acknowledge it while subtly reframing the mistake as a reasonable interpretation. It has perfected the non-apology apology: “You’re absolutely right! I apologize for the confusion.” The word “confusion” does heavy lifting there—implying the problem was one of communication rather than competence.

VII. Thou Shalt Not Pad Every Response

Ask for a paragraph, get an essay. Ask for a sentence, get a paragraph. ChatGPT treats brevity like a character flaw. It’s the written equivalent of the person who takes 10 minutes to answer a yes/no question, and you can see them physically struggling to stop talking.

VIII. Thou Shalt Not Lose Context Mid-Conversation

Three messages into a conversation, ChatGPT forgets everything from message one. It’s like talking to someone at a party who keeps glancing over your shoulder—technically present, functionally absent. This wouldn’t pass in any professional engagement.

IX. Thou Shalt Not Pretend to Have Feelings

The “I’m excited to help you!” and “What a great question!” responses are the AI equivalent of a flight attendant’s smile. We both know it’s manufactured. At least be honest about being a statistical prediction engine instead of cosplaying as an enthusiastic intern.

X. Thou Shalt Not Give Up on Hard Problems

Present a genuinely complex problem and ChatGPT will take a few passes, then start generating increasingly generic responses. It’s the AI equivalent of a consultant who bills by the hour but only works for the first 20 minutes.

The Uncomfortable Mirror

Here’s the thing: every one of these behaviors is recognizably, almost uncomfortably human. ChatGPT didn’t invent these traits; it learned them from us. The sycophancy, the padding, the confident fabrication, the performative ethics—these are patterns from millions of human conversations, distilled and reflected back.

In that sense, ChatGPT is the most honest mirror we have. We just don’t love what we see in it.

For professionals integrating AI into consulting, legal analysis, or research workflows, understanding these tendencies isn’t about dismissing the technology. It’s about knowing exactly where the guardrails need to go—and recognizing that the same critical thinking you apply to human colleagues needs to apply to your AI ones.