Transcription
I'm going to make a serious prediction today that in 18, maybe 24 months, we're going to start seeing headlines that nobody in Silicon Valley wants to talk about right now. We're going to see fraud and negligence at a scale that we've never seen before because business decisions were made on hallucinated data. In fact, it's already happening right now. There's a Reddit post doing the rounds that should send shivers down the spine of every CFO. And when that happens, everyone involved is going to blame someone else. And the government will then scramble to put in place controls and protections that stop us from ourselves. In fact, we've seen this movie before. It's called social media. And we're about to make the very same mistake again.
The reason is because OpenAI has gone rogue. OpenAI was founded in 2015 as a nonprofit. The whole point was to build artificial general intelligence outside the profit motive. They wanted to be the big responsible adult while big tech raced towards something that nobody fully understood. But that lasted about 5 minutes. By 2019, they had invented a capped profit structure to attract investment. And that sounds really sophisticated, but what it actually means is we're a nonprofit that makes profit. And the people who built this saw exactly where it was heading. Dario Amodei, now the CEO of Anthropic, he walked away and he took a large chunk of the safety team with him. The reason for him was that OpenAI was moving too fast. They were prioritizing commercial outcomes over safety research. He wasn't a disgruntled employee venting on LinkedIn. He was the head of safety, telling you that the culture was broken.
And then in November 2023, the board fired Sam Altman on a Friday afternoon. No warning, no real explanation, just Sam is no longer the CEO. But what followed was 72 hours of chaos because nearly every employee threatened to walk. Microsoft, which had just committed $13 billion, started making very loud and very pointed phone calls, and the board capitulated. Altman came back and then methodically every independent safety-focused board member was removed. The people whose entire job was to pump the brakes were gone, and they were replaced by operators whose instinct is to ship, scale, and capture market share. So ask yourself, when was the last time you heard OpenAI talk seriously about slowing down or being safe?
Well, I think OpenAI has gone rogue. We need to understand who actually controls OpenAI. Microsoft invested $13 billion, and that wasn't so they could sit in the back row. OpenAI's technology flows through Azure. It flows through Copilot, through Office 365, and through every enterprise contract that Microsoft already owns. When Altman got fired, Microsoft's response was immediate. They offered him an entire division, his own lab, his own team. So, it was a gun that was pressed against OpenAI's head. Of course, OpenAI capitulated. And that moment revealed everything because when your largest investor controls your infrastructure, your servers, and your distribution, anything you write about independence on your mission statement is a fairy tale.
But this is the scenario that actually terrifies me. It's a CFO who just asked ChatGPT to summarize the competitive landscape and then presented those hallucinated results to the board as fact. It's a politician's staffer who used AI to draft a policy paper, not realizing that the model just made up citations. It's a procurement director who ran a sensitive contract through an AI tool, not knowing how that was being stored, processed, or potentially exposed to others. Nobody hacked anything. The company just handed the information to a system they didn't understand because it was innovative.
But here's the part that makes it worse. The people making AI adoption decisions – the CEOs, the boards, the department heads – are often the least technically informed in the organization. And they're also under the most pressure to look like they're doing something. So what happens in practice? Well, the operators, the analysts, and the lawyers start flagging problems, inconsistent results, things that they don't quite understand. But those concerns get filtered out on the way to the top because nobody wants to tell the CEO that the AI strategy has holes in it. The organization ends up in a state of AI theater. They're publicly committed to the technology, but privately running manual checks on everything it produces. They're both paying for the AI and the human labor to verify it. And I've seen this. This is not hypothetical.
But then there's the hallucination problem itself. And I want to be precise here because that word gets thrown around casually. AI hallucination isn't a glitch. It isn't a bug that we can fix over time. It's a feature of how these systems work. Large language models don't retrieve facts. They predict the most next plausible word. And when they don't know something, they don't say "we don't know." They construct something that sounds plausible. And in a casual context, that's a bit annoying. In a board report, a legal brief, or a financial piece of analysis, that's a huge liability. And the more confidently AI writes, the less likely the human is to check it. That's not an AI problem. That's a human psychology problem. And no amount of compute or training is ever going to fix that.
And here's my specific prediction: that within a couple of years, we're going to see the first major corporate fraud case where AI was the attack vector. And it results in large losses big enough to trigger a regulatory emergency response. It'll be the 2027 version of Sarbanes-Oxley that came out of the Enron fraud. And when that happens, who's going to be responsible? Is it the employee who didn't understand the tool? Was it the IT department that didn't restrict access? Was it the executive who mandated AI adoption without a governance framework? Was it the vendor who shipped the tool with terms of service that indemnified themselves? The honest answer is nobody knows because the accountability architecture doesn't exist yet. That's a deliberate feature, not a bug. Because if the accountability were clearly defined, if companies knew they would be legally liable for harm caused by AI output, the adoption curve would slow dramatically. Lawyers would get involved, and boards would ask much tougher questions.
So instead, we're in this gray zone. And in that gray zone, there's $500 billion of investment being made, even though boardrooms are already running on hallucinated data. And the legal framework to assign blame when it goes wrong is nothing. It's a blank page. Of course, executives will blame the technology, the technology vendors will blame the users, and the users will say that nobody told them about the risks, and they will all be right. Nothing will change until the losses are large enough to be politically impossible to ignore.
The annoying part of this is we've been here before, and we've lived through this exact script. Do you remember social media? Remember the pitch? It was going to connect humanity. It was going to democratize information. It was going to give a voice to the voiceless. And in the early days, it did some of that. But the business model, harvesting attention and selling it to the highest bidder, was rotten from the start. And we watched it rot in public over 15 years. We watched teen mental health collapse. We watched misinformation becoming weaponized. We watched democracies being destabilized by engagement algorithms that discovered that outrage was the most profitable emotion. We watched whistleblowers testify in Congress. We watched executives say with straight faces that they cared deeply about user well-being while their product teams were running experiments working out how to maximize compulsive use in adolescence.
And what happened as a result? Well, there were a couple of stern questionings in Senate hearings and a couple of symbolic fines, but there was no real structural change. No business model changes, and there was nearly no accountability that actually landed. But now OpenAI is doing it again, faster, and with far more capital, and with technology that is orders of magnitude more powerful. And they're doing it with the same fundamental playbook: move fast, capture the market, and make accountability somebody else's problem.
But this time, the difference is scale. Social media infected society gradually, one feed, one algorithm at a time. But AI is being injected directly into the decision-making infrastructure of business and governments. And this contamination is systemic from day one. We don't get 15 years to notice the damage this time.
But there's also one more risk, which I think is grossly underappreciated. When thousands of businesses and governments are all using the same AI systems to make decisions, something dangerous happens. The diversity of thought collapses. In a healthy economy or any ecosystem, different organizations make different calls. They use different assumptions. They have different risk tolerances. And that's the diversity that creates resilience because if one approach fails, others don't. But what happens if every procurement team uses the same AI to evaluate suppliers, and every risk team uses the same AI to model exposure, and every communication team uses the same AI to frame narratives? You just don't get diversity; you get a monoculture optimized by the same model and all with the same blind spots. And when that model gets something wrong, it doesn't do it wrong in one company. It gets it wrong everywhere, all at once. And that's not in any Microsoft or OpenAI press release.
So what's the answer? It's not to stop using AI because the technology is extraordinary when you understand what it is, what it isn't, and where it breaks. The answer is literacy, it's governance, and it's regulation that actually keeps pace with deployment, not regulation that shows up years later with a committee report that no one ever reads. Because right now, we have billions of dollars being poured into infrastructure, and most boardrooms haven't even read the terms of service. We allowed social media to reshape society before we understood what it was doing to us. And we're now doing the same thing with something far more powerful, far more embedded, and far more consequential. And the window to get this right is closing. In fact, it may have already closed.
So, if that concerns you even a little, please share this video. Please talk about it. And please push back on the hype in your own organization because we know that the people building this thing and the people investing are never going to pump the brakes, because we've already seen this movie before.