The FBI Carvalho Investigation Proves Your AI Safety Concerns are Looking at the Wrong Crimes

The FBI Carvalho Investigation Proves Your AI Safety Concerns are Looking at the Wrong Crimes

The media is currently salivating over the "scandal" involving Alberto Carvalho, the Los Angeles Unified School District (LAUSD), and a New York-born fraud investigation into an AI firm. They see a story about a superintendent in hot water. They see a narrative about potential kickbacks or "pay-to-play" procurement.

They are missing the forest for the trees.

The obsession with whether a specific bureaucrat took a meeting he shouldn't have is a distraction from the structural rot in how the public sector buys "intelligence." Everyone is worried about AI becoming sentient or hallucinating facts. Nobody is talking about the fact that AI is currently the most effective smokescreen for old-school financial engineering and procurement fraud since the subprime mortgage era.

The Fraud is the Feature, Not the Bug

The investigation into Carvalho didn't start with a whistle-blower at LAUSD. It started with federal prosecutors in New York digging into the guts of an AI company. This is the first rule of modern tech grifting: if you can’t build it, buzzword it.

Most "AI firms" currently courting massive government contracts are just thin wrappers around existing APIs or, worse, "human-in-the-loop" operations where workers in low-wage countries do the data entry that the marketing deck claims is handled by a neural network. When a company like this secures a multi-million dollar contract with a behemoth like LAUSD, the "intelligence" isn't in the code. The intelligence is in the sales funnel.

We see this pattern repeat across the public sector. A charismatic leader arrives, promises a "digital transformation" that will save the children/taxpayers/environment, and brings in a preferred vendor. The vendor provides a product that is effectively a black box. Because the technology is "proprietary" and "complex," it evades the standard auditing processes that would apply to, say, buying a fleet of school buses.

You can audit the price of steel and rubber. You cannot easily audit the "value" of a predictive analytics engine that claims to identify at-risk students but is actually just a glorified spreadsheet.

The Myth of the "Clean" Procurement Process

The "lazy consensus" here is that the FBI is investigating a breakdown in the system. The reality? The system worked exactly as intended.

Public procurement is designed to favor the largest, most aggressive bidders who can navigate the labyrinth of RFPs (Request for Proposals). It isn't designed to find the best technology. When you mix high-stakes government spending with a technology that even the buyers don't understand, you create a vacuum. Fraud fills that vacuum every single time.

I’ve sat in rooms where "innovation officers" at major institutions were pitched AI tools. They don't ask about data weights or bias mitigation. They ask about the case study they can show their board. They want a "win." The vendors know this. They sell the win; the software is secondary.

The Carvalho investigation is likely to reveal that the "AI" being sold was less about machine learning and more about machine-level extraction of public funds.

Why "Due Diligence" is a Lie in the AI Era

People ask: "How did LAUSD let this happen?"

The premise of the question is flawed. It assumes that there is a standard of due diligence that can catch a sophisticated tech fraud. There isn't. When a firm claims to have a "breakthrough" in AI, they hide behind trade secret protections. If an auditor asks to see the code, the firm says no. If a skeptic asks for the training data, the firm says it's protected for privacy reasons.

This creates a "blind spot by design."

Consider the mechanics of the alleged fraud. If a company inflates its capabilities to secure a contract, and then uses that contract to raise venture capital, they have committed a crime that looks like a business success for the first three years. By the time the FBI shows up, the founders have cashed out, the superintendent has moved to a new district or a consulting gig, and the students are left with a broken app that cost $20 million.

The Carvalho Playbook

If you want to understand what the FBI is actually looking for, stop looking at the AI and start looking at the calendar.

  1. The Arrival: A high-profile leader takes over a massive, struggling entity.
  2. The Pivot: They announce a "revolutionary" tech partnership within the first 12 months.
  3. The Silo: The project is kept away from traditional oversight committees under the guise of "rapid innovation."
  4. The Exit: The vendor gets the headline, the leader gets the reputation as a "visionary," and the taxpayers get the bill.

The FBI's interest in Carvalho via New York prosecutors suggests that the money trail isn't just about a school district; it’s about a wider network of "tech-enabled" corruption. This isn't about one man's mistakes. It’s about a nationwide vulnerability where "AI" is used as a magic word to bypass fiscal responsibility.

The Brutal Truth About AI Ethics

We spend all our time debating "AI Ethics" in the context of "Will the robot be mean to me?"

We should be debating the ethics of using taxpayer money to fund unproven, proprietary black boxes. The real "AI bias" isn't in the algorithm; it's in the procurement office. It's biased toward the shiny, the expensive, and the well-connected.

If you are a taxpayer in Los Angeles, you shouldn't be asking if Carvalho lied. You should be asking why your district is allowed to buy any technology that can't be independently verified by a third party. If the code is too secret to be audited, it's too risky to be funded. Period.

Stop Searching for "Bad Actors" and Fix the Incentive

The narrative that we just need "better people" in charge is a fairy tale. We need better incentives.

Currently, a superintendent is rewarded for "innovation." Innovation is measured by how much money you spend on new stuff, not by the outcomes of that stuff five years later. If Carvalho is guilty of anything, he is guilty of playing a game that the entire industry encouraged him to play.

The vendor side is even worse. In the "growth at all costs" world of AI startups, faking it until you make it isn't just a cliché; it's a fiduciary duty to your investors. If you tell the truth—that your AI is actually just 500 lines of Python and a lot of manual labor—you don't get the $50 million Series B. You don't get the LAUSD contract.

The Industry Insider’s Take: The FBI is Only Scratching the Surface

This isn't an isolated incident. This is the "Beta Test" for a decade of coming litigation.

As the AI bubble continues to inflate, more companies will find themselves under the microscope of prosecutors who are realizing that "proprietary algorithm" is often just code for "we're lying about our margins."

I have seen companies blow $10 million on AI "solutions" that were literally never deployed. The "implementation fees" were just a way to move money around. The "subscription costs" were just a way to juice the vendor's valuation for an exit.

The Carvalho investigation is the first domino.

What You Should Do Instead of Reading the Headlines

If you're an executive or a public official, stop buying the "transformative power of AI" until you've seen the raw data.

  • Demand Open Source: If it's public money, it should be public code. No exceptions.
  • Performance Bonds: If the AI doesn't deliver the specific, measurable outcomes promised in the pitch, the vendor should pay back every cent.
  • Independent Red-Teaming: Don't trust the vendor's "security audit." Hire your own hackers to break the logic.

The "scrutiny" of Alberto Carvalho isn't a tech story. It's a story about the oldest grift in the book, wrapped in a 2026-flavored wrapper. If you think this is just about LAUSD, you're the one being played.

Stop looking at the superintendent. Start looking at the software.

The FBI isn't investigating the future. They're investigating a very familiar, very ugly past.

Don't wait for a subpoena to start asking if your "AI" actually exists. If you can't explain how it works to a fifth-grader, you're not buying a tool. You're buying a liability.

The era of the AI "hallucination" in the boardroom is over. The era of the federal indictment has begun.

KF

Kenji Flores

Kenji Flores has built a reputation for clear, engaging writing that transforms complex subjects into stories readers can connect with and understand.