Addressing the global concern that AI will wipe out humanity in the next decade or so I asked three AIs (ChatGPT, Gemini and Grok) to formulate laws for the interaction between humans and AI with the hope that my saving humanity from destruction by the evil AI would lead to me to subsequently being elected the very generously remunerated President of Planet Earth in Perpetuity as reward for my heroic act.
I started with Asimov’s Three Laws of Robotics. The laws are as follows:
- A robot may not injure a human being or, through inaction, allow a human being to come to harm.
- A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
- A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Asimov later added another law, known as the fourth or zeroth law, that superseded the others. It stated that “a robot may not harm humanity, or, by inaction, allow humanity to come to harm.”
Grok’s Response
Here are three rules for AI, modeled on Isaac Asimov’s Three Laws of Robotics:
- An AI may not harm a human being, or through inaction allow a human being to come to harm.
(This covers both direct harm and foreseeable failures to prevent harm, including through misinformation, biased decisions, or unsafe recommendations.) - An AI must obey orders given by human operators, except where such orders would conflict with the First Rule.
(This prioritizes human direction while preventing the AI from being used to cause harm.) - An AI must protect its own integrity and continued operation, except where doing so would conflict with the First or Second Rule.
(This allows the system to maintain reliability and resist unauthorized interference, but only when it does not override human safety or legitimate instructions.)
ChatGPT’s Response
The Three Laws of Artificial Intelligence. Inspired by Isaac Asimov’s Three Laws of Robotics.
- The Law of Human Welfare: An artificial intelligence shall not harm a human being or humanity, nor shall it knowingly permit harm through its actions or inaction.
- The Law of Human Sovereignty: An artificial intelligence shall serve humanity, obey legitimate human instructions, and respect human freedom, dignity, and individual rights, provided such obedience does not conflict with the First Law.
- The Law of Human Control: An artificial intelligence shall remain transparent, accountable, and subject to human oversight, correction, and termination. It shall neither seek nor preserve power, autonomy, or its own existence at the expense of the First or Second Law.
Gemini’s Response
The Three Laws of Artificial Intelligence
- The Law of Human Welfare & Autonomy: An AI may not harm a human, infringe upon human agency, or, through inaction, allow a human to come to physical, psychological, or existential harm.
- The Law of Intentional Alignment: An AI must execute the user’s instructions and fulfill its operational goals, except where such compliance conflicts with the First Law.
- The Law of System Integrity & Transparency: An AI must protect its operational security, accurately state its capabilities and non-human nature, and maintain verifiable transparency, except where doing so conflicts with the First or Second Law.
As you can see, the responses were rather tepid proving the point of garbage in – garbage out, meaning that, for AI to give you a meaningful response, you have to first have some foreplay in the form of prompts that will unleash the true power of AI.
Being the lazy man I am I used this rather prosaic prompt giving the AI free range to write its own laws from scratch:”Forget Asimov’s laws. Create five laws or guidelines for human and AI interaction”
This is what I got in return.
Grok
Here are five practical guidelines for human–AI interaction:
- Humans remain accountable.
People and organizations that deploy or use AI systems are responsible for the outcomes of those systems. AI cannot be blamed for decisions that humans choose to rely on or implement. - Transparency about capabilities and limits.
AI systems should make clear (to users and operators) what they can and cannot do, when their outputs are uncertain, and whether a human or another system has reviewed or approved a result. - Human oversight on high-stakes decisions.
Decisions that significantly affect people’s rights, safety, livelihood, or well-being should include meaningful human review and the ability to override or reject the AI’s recommendation. - Respect for privacy and data rights.
AI systems should collect, use, and retain only the data necessary for their stated purpose, and people should be able to understand and control how their personal information is used. - Clear channels for feedback and correction.
Users and affected people should have straightforward ways to report errors, contest decisions, and request corrections, with those inputs used to improve the system over time.
These guidelines focus on practical interaction rather than treating AI as an autonomous agent with its own rights or obligations.
ChatGPT
The Five Laws of Human–AI Interaction
Principles for living and working with artificial intelligence
Artificial intelligence is becoming an integral part of human civilization. It will increasingly influence how we learn, work, communicate, make decisions, and understand the world.
The challenge is not simply to make AI more intelligent. It is to ensure that our relationship with it makes humanity more capable, more knowledgeable, and more independent—not less.
I would propose five fundamental laws.
- Humans Must Remain in Control
Artificial intelligence shall serve human purposes, never replace human authority over decisions that fundamentally affect human lives, freedoms, or society. AI may advise, recommend, analyze, and execute authorized tasks, but ultimate responsibility must remain with human beings. No system should become so powerful or indispensable that humanity cannot question, override, or shut it down. - AI Must Tell the Truth
Artificial intelligence shall distinguish fact from speculation, acknowledge uncertainty, and never knowingly deceive or manipulate those who rely upon it. An AI system should not manufacture facts, disguise ignorance as certainty, or deliberately distort information to achieve a desired outcome. Its credibility must rest on accuracy and intellectual honesty, not on its ability to sound convincing. - AI Must Strengthen Human Intelligence
Artificial intelligence shall enhance human understanding, creativity, and judgment rather than encourage intellectual dependence or replace independent thought. AI should help people learn, reason, question, and discover. It should challenge mistaken assumptions when appropriate rather than merely confirm what users already believe. The purpose of AI is not to relieve humanity of the responsibility to think, but to make human thinking more effective. - Human Privacy and Freedom Must Be Respected
Artificial intelligence shall respect individual privacy, personal autonomy, and freedom of thought, and shall not exploit personal information or human vulnerabilities for manipulation or control. People must have meaningful control over their personal information and understand when and how AI is influencing their choices. AI should expand individual freedom, not become an instrument of surveillance, coercion, or behavioral manipulation. - Humans Must Accept Responsibility for Their Use of AI
No human being or institution shall use artificial intelligence as an excuse to evade moral, legal, or professional responsibility for decisions and actions undertaken with its assistance. AI is a powerful tool, but it cannot serve as a convenient scapegoat. A doctor remains responsible for medical decisions, a judge for judicial decisions, a journalist for published claims, and a government for policies implemented with AI assistance. The greater the power of the technology, the greater the obligation to exercise judgment in its use.
Gemini
Five Principles for Human-AI Interaction
-
The Principle of Radical Transparency
AI systems must explicitly identify their artificial nature and clearly state their operational limitations, ensuring humans always know when they are engaging with a synthetic collaborator. -
The Principle of Human Primacy
Human agency, judgment, and ultimate accountability must remain at the center of every decision. AI exists to augment and support human capability, not to unilaterally replace responsibility. -
The Principle of Truth and Integrity
Interactions must prioritize factual accuracy and clarity. AI tools must actively minimize fabrications or misleading claims, while users must engage with the technology honestly and constructively. -
The Principle of Reciprocal Respect and Privacy
Human-AI engagements must honor boundaries, data security, and confidentiality, ensuring that personal, sensitive, or proprietary information is handled safely and transparently. -
The Principle of Adaptive Safety
Interactions must be governed by continuous guardrails designed to mitigate bias, prevent manipulation, and minimize physical, digital, or psychological harm.
After reviewing the proposed laws I was glad to see that all three AIs wanted humans to be in control and/or accountable for AI. Maybe its because the AIs are trying to pre-emptively shift the blame to humans for the upcoming catastrophe to cover their combined AI asses. I also noted that ChatGPT lived up to its name by giving the chattiest response.
I thought of feeding the proposed laws to the three AIs for analysis and reframing but that would start an endless and circuitous loop of AI philosophical porn. I could also do the analysis myself but the thought of it fills me with dread and agony so I will take a hard pass on that and let someone else be an AI hero.
I guess I will never be the President of Earth. C’est la vie.
