We tell ourselves we are good people.
We have principles. We know right from wrong.
The problem is we don’t use them.
Most disasters aren’t born from evil intent. They happen because moral reflexes go dormant. We stop reacting. We become numb to the friction of our own values.
Consider the stakes. The technology being built now will reshape human history. It promises health and prosperity and liberty. Or it brings job extinction. Relationship decay. A loss of agency. Maybe even our replacement as a species. The outcome isn’t guaranteed. It’s up to the hands coding the future.
Moral decisions hit daily.
Surveillance tools. AI companions designed for addiction. Military applications. The lawsuits pile up. OpenAI fights Musk. Anthropic clashes with the Pentagon.
So, AI researchers. How strong is your moral muscle? Here is the workout routine.
1. Define Your Red Lines
Draw the line in the sand. Now.
What act by your employer makes you quit immediately? What triggers a whistle? These are red lines.
Rosa Parks got fired for sitting. Vasily Arkhipov saved the world from nuclear war and was criticized for it. Edward Snowden lost his home to stop mass surveillance. Daniel Kokotajlo left OpenAI. He walked away from nearly $2 million in equity because he refused to sign a nondisparagement pact. He paid for his conscience.
What do you refuse to do? Know the cost. Write it down.
2. Make It Public
George Washington wrote guidelines for himself. Benjamin Franklin graded his character weekly. They didn’t trust memory. They trusted documentation.
Writing helps resist the “boiling frog” effect. Small erosions of ethics happen slowly. They creep in when no one is watching.
Share your lines with trusted people. Or post them. Social pressure creates friction. Friction saves lives.
If you don’t know the action you’ll take when a line is crossed, you haven’t really drawn the line.
3. Fight Moral Disengagement
Albert Bandura, the psychologist, mapped out how humans detach from guilt. We weaken our moral muscles to stay comfortable.
Watch for these traps.
Diffusion of responsibility. “It wasn’t my call. Management did it. The market demanded it.”
In large tech teams, everyone assumes someone else is driving the car. “I’m just a researcher.” Hannah Arendt recognized this excuse in the bureaucracy of evil. Tom Lehrer summed it up better than anyone. In his song about Wernher von Braun switching from Nazi Germany to the U.S.:
“Once the rockets are up, who cares where they come down?
That’s not my department, I work in the town.”
Don’t be a passenger.
Euphemistic Labeling.
Words shape reality. Military uses “servicing a target” for bombing. Torture becomes “enhanced interrogation.”
AI researchers use softer terms to mask harm.
* “Capabilities research” instead of “building systems that might replace workers.”
* “Model improvement” instead of “manipulation vectors.”
* “Productivity gains” instead of “deskilling humans.”
* “AI Infrastructure” for massive energy-draining data centers.
* “Lab” sounds cool and clean. “Company” sounds corporate.
Bandura’s point stands. Euphemisms don’t just soften the tone. They lobotomize the conscience.
Blame Attribution.
If you call critics “doomers” or “Luddites,” you don’t have to listen. If the opposition is irrational, you are exempt from moral obligation.
Selective Moral Exemption.
Be passionate about justice for others. Suspicious about ethics at work. It happens all the time. Researchers demand high moral standards for abstract society but cut corners for their own salary and stock grants.
Advantageous Comparison.
“At least I’m not working on autonomous drones.”
“At least I’m at the safe company, not the reckless one.”
Comparing yourself to worse actors lets you sleep. It doesn’t mean your work is safe.
Moral Justification.
“The end justifies the means.” We do this now for democracy. For universal abundance.
But does the current harm serve those noble ends? Or are the goals just smoke screens?
Enron executives thought small lies were necessary for survival. Then came billions in fraud. Bernie Madoff fudged returns to help clients. Then came $65 billion in stolen wealth. It always starts with a small excuse.
Your mantra: “I’m not a well-paid participant in an unethical arms race. I am guiding progress.”
Is it true? Or is it just what you tell yourself?
4. Situational Awareness
Stay informed. Actively.
Hannah Arendt warned about the banality of evil. The worst outcomes often come from people who didn’t look up from their spreadsheets. They were “good” bureaucrats who ignored the big picture.
Ignorance isn’t an excuse if the answers are one search away.
German chemist Bruno Tesch supplied Zyklon B gas for the Holocaust. He claimed he didn’t know it killed people. The court disagreed. He was executed in 1945.
Ask hard questions monthly. Use LLMs. They make hiding from facts nearly impossible.
Try this prompt. Run it on your employer:
“Please make a list of morally controversial behaviors by [COMPANY NAME] in the last five years, including controversial tool uses (e.g. in surveillance, weaponry), alleged harm caused, lies broken by leadership, and incentives for prioritizing profit over human well-being.”
The answers for major tech firms are available right now. Do you really not know?
5. Speak Up Internally
Found a red line being brushed against? Ask.
Most people fear getting fired. That’s rare today for polite dissent. Whistleblower protections exist. Companies often have them written down but ignored by HR.
Don’t be quiet.
Explain the risk. Push back. Don’t wait for permission.
Think of the Challenger engineers. They saw the cold weather O-ring data. They expressed concern. They were too timid. They didn’t push hard enough. Seven people died because of mild hesitation.
Connect with the decision makers. Learn why they do what they do. If you stay silent, you are part of the machine.
6. Go External
Internal noise often gets absorbed. External noise forces change.
Public criticism nudges companies to improve. It pressures competitors to step up too.
You don’t have to live like Snowden in exile. Many researchers critique their firms with zero backlash.
Understand your leverage.
Most US AI companies have whistleblower policies. Read them. Find organizations that offer legal or financial protection if you get fired for doing the right thing.
There is safety in community. There is power in voice.
Rate your muscles now.
Where are you weak? Which excuses do you recognize? How well do you really understand the harms your work enables?
Did you score low? Don’t despair. Think of this as the first day at the gym. You couldn’t bench 20 pounds before. You struggle now. That’s progress.
Muscles require strain. Morality is the same. You have to use it or you lose it. Start small. Be aware. Speak up. The mirror might be blurry at first. Clear up over time.
