I spent half a decade and probably a couple million dollars chasing a ghost: the perfect AI ethics framework. It was a noble quest, or so I thought. After two successful exits with RemoteTeam and MovieLaLa, and investing in over 200 companies including some of the biggest names in AI like Anthropic and OpenAI, I figured I had a responsibility to get this right. I was wrong. Not about the responsibility, but about how to discharge it. My journey into the world of AI ethics was my biggest professional failure, and it taught me a lesson that I want to share with you, so you don't have to waste the same five years I did.
The Early Days: Chasing the Dragon of Complexity
Back in 2019, AI ethics was the new hotness. Every conference had a track on it, every VC was asking about it, and every big tech company was publishing their own set of lofty principles. I dove in headfirst. I hired consultants who came with impressive slide decks and even more impressive invoices. I remember one consultant, a former academic with a PhD in philosophy, who charged us $20,000 for a two-day workshop that produced a list of five "core values." I’m still not entirely sure what they were. I think one of them was "humanity," which is about as useful as a screen door on a submarine.
We workshopped, we brainstormed, we wordsmithed. The result? An 80-page "Principled AI" document that was a masterpiece of academic jargon. It had matrices, scoring systems, stakeholder maps, and a color-coded risk assessment framework. It was beautiful. And it was completely useless. That document, which we spent months creating, sat in a shared drive, collecting digital dust. I’d bring it up in meetings, and my engineers would nod politely, their eyes glazing over. Then they’d go back to their desks and ship code. The fancy framework had zero impact on our product development process. It was a corporate artifact, a shield to show investors and journalists that we were "thinking about the hard problems." But we weren't. We were just talking about them.
I honestly had no idea what I was doing. I was just following the herd. I saw what Google and Microsoft were doing, and I thought that was the only way. I was so focused on the appearance of being ethical that I lost sight of what it actually means to build responsible technology. I was treating ethics as a PR problem, not a product problem. And that was a huge mistake. It was all hat and no cattle. We were performing ethics, not practicing it. We had a Chief Ethics Officer who had never shipped a line of code in her life. She was a lawyer by training, and her main job was to make sure we didn't get sued. That's not ethics. That's risk management.
The Turning Point: The MovieLaLa Debacle
The wake-up call came in the form of a feature we launched at MovieLaLa. It was a recommendation engine, designed to help users discover new films. On the surface, it was a success. Engagement went up, and users were watching more movies. But then we started hearing whispers. Users were complaining that they were only seeing the same kinds of movies, that their recommendations felt… stale. We had inadvertently created a massive filter bubble. The algorithm, in its quest to optimize for engagement, was reinforcing users' existing biases, and in the process, making our product less useful.
I remember the day it all came to a head. We were in a board meeting, and one of our investors, a sharp woman who had been a product manager at Netflix for years, pulled up her MovieLaLa account. Her entire feed was just romantic comedies. "I watched one romantic comedy three weeks ago," she said, "and now your app thinks that's all I want to watch. This is broken." She was right. And I had no good answer for her. I mumbled something about "personalization" and "user preferences," but we both knew it was bullshit.
All our beautiful frameworks were useless in that moment. We were flying blind, and the fancy document couldn't help us. We had to scramble, pulling engineers off other projects to fix the recommendation engine. It was a costly, embarrassing mess. And it was then that I realized the problem wasn't a lack of principles, but a lack of process. Our 80-page document was full of good intentions, but it didn't give my team a single actionable step to take when they were designing a new feature. It was like having a fire extinguisher manual, but no fire extinguisher. We had a lot of theories about fire safety, but we had no way to put out the fire that was burning down our product.
The Shift: From Frameworks to a Simple, Actionable Checklist
After the MovieLaLa debacle, I threw out the 80-page document. I was done with frameworks. I wanted something that an engineer could use in a 15-minute meeting, something that would spark a real conversation, not just a series of polite nods. So, I came up with a simple, one-page checklist. I wanted a simple set of questions that would force a real conversation. It boiled down to a couple of key things.
First, we always ask: Why are we building this? This question forces you to articulate the user need. Is this a feature that users are actually asking for, or is it just a cool piece of tech that we want to build? This is the first line of defense against tech-for-tech's-sake. It’s amazing how many bad ideas you can kill just by asking this one simple question. I’ve seen entire roadmaps get thrown out because nobody could give a good answer to this question. It's a question that every founder should be asking themselves every single day.
Second, and this is the most important one, we ask: Why could this go wrong? This is where the magic happens. I ask my teams to brainstorm all the ways a feature could be misused, or have unintended negative consequences. We call it "red-teaming for ethics." What are the evil use cases? How could this be used to discriminate, to spread misinformation, to harm people? This is often the most uncomfortable part of the conversation, but it's also the most important. It's about anticipating the negative externalities of your work. We do this for every single feature, no matter how small. It’s a non-negotiable part of our process. It's not a one-time thing. It's a muscle that you have to build and maintain.
And finally, we ask ourselves: Why can't we fix it later? This question is designed to combat the "we'll fix it in post" mentality that is so pervasive in Silicon Valley. It forces a discussion about proactive mitigation versus reactive cleanup. Sometimes, you can fix things later. But often, by the time you realize something has gone wrong, the damage is already done. Your reputation is shot, your users have left, and you’re facing a mountain of technical debt. It’s almost always cheaper to fix things before you ship them. I tell my founders that a dollar spent on prevention is worth a hundred dollars spent on the cure.
This simple checklist has been more effective than any 80-page framework. When I was looking at the early pitch from Anthropic, we spent more time on the "Why could this go wrong?" than on the Total Addressable Market. That's the conversation that matters. It’s the same reason I wrote about how to build a great company culture. It’s about the daily practice, not the posters on the wall.
Regulation is Coming, and It Won’t Save You
I know what some of you are thinking. "This is all well and good, Sahin, but what about the EU AI Act? What about regulation?" Look, I’m not against regulation. The EU AI Act is a well-intentioned, and in some ways, necessary piece of legislation. But it will not save you. It will become another complex framework that companies will try to "hack" or comply with on paper only. It’s like GDPR. Every website now has a cookie banner, but has it really changed the underlying data economy? Not in a meaningful way.
I was in Brussels last year, meeting with some of the folks who were drafting the AI Act. They are smart, dedicated people. But they are not builders. They don't understand the pressures of a startup, the need to ship code every day. They are trying to solve a dynamic problem with a static solution. It's like trying to build a car by writing a book about cars. It just doesn't work. The world of AI is moving too fast for regulation to keep up. By the time the AI Act is fully implemented, it will already be obsolete.
Real change doesn't come from a government mandate. It comes from a cultural shift within your organization. It comes from making these conversations a part of your weekly routine. It comes from empowering your engineers to raise red flags, and from rewarding them when they do. Internal culture and simple, repeatable processes are more powerful than any external regulation. If you’re waiting for the government to tell you how to build ethical AI, you’re already behind. You're playing defense when you should be playing offense.
What I Do Now
Now, when I advise my portfolio companies, all 200+ of them, I don't ask to see their AI ethics framework. I don't care about their fancy documents. I have a very different conversation with them. A founder came to me last week with a new AI-powered hiring tool. I didn’t ask to see their ethics framework. I asked them: "Tell me three ways someone could use this to discriminate, even accidentally." The silence was telling. It was the beginning of a real conversation, a conversation that will have a far greater impact on their product than any 80-page document ever could.
We talked for an hour about how they could build in safeguards, how they could monitor for bias, how they could give users more control over the process. We talked about the importance of transparency, and how they could explain to users how the algorithm works. We talked about the need for a human in the loop, and how they could design a system that empowers recruiters, rather than replacing them. It was a messy, difficult conversation. But it was a productive one. And it was a conversation that would have never happened if I had just asked to see their ethics framework. I didn't give them a solution. I gave them a process for finding their own solution.
It all comes down to the principles I outlined in my book, Becoming Top 1%. It’s about focusing on the things that actually matter, and having the courage to ignore the things that don’t. It’s about people and process, not paperwork. It’s about having the humility to admit that you don’t have all the answers, and the wisdom to ask the right questions. It's about being more of a gardener than an architect. You can't design a perfect system from the top down. You have to cultivate it from the ground up.
Conclusion
I wasted five years chasing the ghost of the perfect AI ethics framework. I thought I could solve a human problem with a technical solution. I was wrong. The answer isn't a more complex framework. It's a simpler conversation. It's asking the uncomfortable questions, and making the time to have the hard conversations. That's it. That's the secret.
I wasted five years so you don't have to. Now go build something that doesn't break the world.
Frequently Asked Questions
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.