Forget the EU AI Act. It’s already a relic.
I know that sounds crazy. People are still scrambling to understand its 800 pages. But I’ve seen this movie before. I was there for the dot-com bust that vaporized billions in paper wealth overnight. I was there when the iPhone created a new universe of mobile-first companies, and my startup, RemoteTeam, was one of them before being acquired by Gusto. I saw the power of community-driven content with MovieLaLa, which was later acquired by Gfycat. Now, I’m an angel investor in over 200 companies, including some of the foundational players in this new AI world like Anthropic, OpenAI, Scale AI, and Hugging Face.
Living through these cycles teaches you to spot patterns. And the pattern I see now is that the current approach to AI regulation is about to become obsolete. It’s a well-intentioned but clumsy attempt to put a leash on a rocket ship. By 2027, the conversation will have completely changed. The rulebook will be rewritten.
So, if you want to know where the puck is going, not where it is, here are my three bold predictions for the AI regulatory landscape in 2027. This isn’t academic theory. This is a playbook for survival and success in the next phase of the AI revolution.
Prediction 1: The End of "General-Purpose AI" as a Legal Shield
Remember the term "information superhighway"? In the 90s, that’s what we called the internet. It was a vague, catch-all phrase that meant everything and nothing. That’s exactly what "general-purpose AI" is today. It’s a convenient label for foundation models like GPT-4 that can do a thousand different things. It’s also a massive legal loophole.
Right now, the builders of these huge models are hiding behind this term. They argue that they can’t be held responsible for all the downstream uses of their technology. They just provide the raw intelligence, and it’s up to the application developers to use it responsibly. That argument is about to fall apart.
By 2027, regulators will have wised up. They’ll realize that regulating "general-purpose AI" is like trying to regulate "general-purpose electricity." It’s pointless. Instead, they will shift their focus to regulating specific, high-risk AI capabilities. Think of it as a surgical strike instead of a blanket ban.
What does this look like in practice? Instead of a single set of rules for all AI, we’ll see specific regulations for:
- AI-powered medical diagnostics: Any AI that suggests a diagnosis will face the same level of scrutiny as a new medical device. The FDA will be all over this.
- AI-driven credit scoring and lending: The Consumer Financial Protection Bureau (CFPB) will demand full transparency on how these models work to prevent discriminatory outcomes.
- AI in hiring and recruitment: The Equal Employment Opportunity Commission (EEOC) will require audits to ensure AI tools aren’t filtering out candidates based on protected characteristics.
I made an early investment in Scale AI for a reason. They understood that the future of AI wasn’t just about bigger models, but about better, more specialized data. They provide the high-quality, human-verified data needed to fine-tune models for specific, mission-critical tasks. That’s where the real value is, and that’s what regulators will ultimately zoom in on. The era of offloading responsibility is over. If your model has the capability to cause harm in a regulated industry, you will be held accountable.
Prediction 2: Mandatory AI "Nutrition Labels" on Every Model
When we were building MovieLaLa, we had to deal with a chaotic mess of movie metadata. Different studios had different formats, rights were a nightmare, and there was no single source of truth. It was a data governance disaster. We are heading for a similar crisis with AI models, and the solution will be what I call "AI Nutrition Labels."
Think about it. You wouldn’t eat food without knowing the ingredients. You wouldn’t take medicine without reading the side effects. Why should we use powerful AI models without knowing what’s inside them?
By 2027, it will be a legal requirement for every commercial AI model to come with a standardized "label" that clearly discloses key information. This won’t be a voluntary best practice; it will be the law. This label will have to include:
- Training Data: What datasets was the model trained on? Were they public, proprietary, or scraped from the web? Was copyrighted material used? Expect a full manifest.
- Known Biases: Every model has biases. The label will have to explicitly state the model’s known limitations and weaknesses, particularly regarding race, gender, and other sensitive attributes.
- Performance Benchmarks: How does the model perform on standardized industry tests for accuracy, fairness, and robustness? These benchmarks will be audited by third parties.
- Energy Consumption and Carbon Footprint: The environmental cost of training and running these massive models is becoming a huge issue. The label will have to disclose the energy consumed and the resulting carbon emissions.
My investment in Hugging Face was a bet on this future. Their entire platform is built around the idea of open and transparent AI. Their model cards were a precursor to this idea—a voluntary effort to document what a model is and how it should be used. But in the next few years, this will move from a community norm to a regulatory mandate. Companies that are already transparent will have a massive head start. Those that operate in secrecy will face a painful reckoning.
Prediction 3: The Rise of the Certified AI Auditor
After the dot-com bubble burst in 2000, the Sarbanes-Oxley Act created a new era of financial accountability. It established strict new rules for accounting firms and corporate governance. We are about to see the AI equivalent of this. Just as we have certified public accountants (CPAs) to audit financial statements, we will have Certified AI Auditors to audit algorithms.
Governments simply don’t have the technical expertise or the manpower to inspect every complex AI system. It’s not feasible. Instead, they will do what they always do: they will outsource the job. They will create a new professional class of certified experts who are licensed to audit AI models for safety, fairness, and compliance.
This will create a whole new industry. These auditors will be the new gatekeepers. You won’t be able to deploy a high-risk AI system without getting it signed off by a certified auditor. This certification will become a prerequisite for getting insurance, securing enterprise contracts, and avoiding massive fines.
This is why I invested in Anthropic. From day one, their entire mission has been about building safe, steerable AI. They have been thinking about the "how" as much as the "what." They are building the technical foundations for AI safety that will make auditing possible. Companies like this, which have baked safety into their DNA, will be the first in line to work with and define the standards for this new profession.
This isn’t just about compliance; it’s about trust. The public is becoming more and more skeptical of AI. A seal of approval from a trusted, independent auditor will become a powerful competitive advantage.
Your One Move for 2027
So what does this all mean for you? You can’t just wait for these regulations to happen. You need to start preparing now. And the single most important move you can make is to start documenting everything.
Treat your AI development process like a crime scene. Document every decision, every dataset, every test, and every trade-off. Create your own internal "nutrition labels" before you are forced to. Build a culture of transparency and accountability within your team.
Because when the regulators come knocking—and they will—the companies that can show their work will be the ones that survive and thrive. The ones that treated their AI models as inscrutable black boxes will be buried in lawsuits and fines.
The next three years will be a mad dash. The AI landscape of 2027 will look nothing like it does today. The winners will be those who see around the corner and build for the world that is coming, not the world that is.
Frequently Asked Questions
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.
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 can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.