Opinion | Real Despots Hijack Artificial Intelligence - The New York Times

The Sinking Ship of 'Progress'

Look, I've seen a lot of hype cycles come and go in this business. Two decades of watching silver bullets turn to lead balloons. Dot-com bubble. Cloud everything. Blockchain for democracy, my ass. Now it’s AI, and the stench of desperation from VCs and politicians alike is frankly nauseating. Everyone's shouting about AGI, about "democratizing intelligence," while the real game is playing out in the shadows. The reality is, the very systems we built, the shiny new toys, they're not just being misused; they're being actively weaponized by the worst among us. And we built the damn weapons.

We’ve been selling a dream, haven’t we? Productivity. Efficiency. A brighter future. Bullshit. What we’ve actually built, piece by piece, is an infrastructure of control, ripe for the taking. Remember the early days, talking about BSS/OSS for better customer experience? Total nonsense. It was always about consolidating power, streamlining operations to cut costs, and yes, gathering data. Mountains of it. Data we swore would be used for good, for personalization, for understanding user behavior. Instead, it became the raw crude for the next generation of digital despots. The juice isn't worth the squeeze, not when the squeeze means handing over the keys to the kingdom.

The architects of our digital future, the ones with the big smiles and even bigger valuations, they talk about ethical AI. They talk about guardrails. Laughable. It’s like designing a super-fast car and then being surprised when someone drives it off a cliff. Or, more accurately, when someone drives it straight into a crowd. The incentives were always perverse. Build it fast, make it scalable, don't worry about the ethics until there's a PR crisis. We ignored the warnings, the small tremors, because the numbers were good. Now, the earthquake is here, and we're all pretending to be surprised. It’s a tragedy, really, and we're all complicit in this grand farce.

The Data Graveyard

Let's talk about data, the supposed fuel of this revolution. We’ve been collecting it like digital hoarders for years. Every click, every purchase, every casual conversation with a smart speaker. We called it "big data." Sounds impressive, right? Most of it's garbage. Rotting in data lakes, poorly labeled, full of bias. But even the garbage, once processed through enough algorithms, can reveal patterns. Patterns that, in the wrong hands, become chains. And those chains are getting stronger. The promise of impartial algorithms? A fairy tale spun by people who never spent a day in the trenches with real-world data, trying to make sense of the glorious mess humans create.

Think about the sheer volume. Petabytes of personal information, stored in data centers that hum with the quiet desperation of a million forgotten secrets. Companies spent fortunes on MPLS networks just to shuttle this stuff around. For what? To feed models that are now being fine-tuned not just for targeted ads, but for targeted oppression. The original sin wasn't just collecting it; it was convincing ourselves that simply having more data equated to more truth, more insight. We forgot that bias in, bias out is a fundamental law of the digital universe. These LLMs, they don’t discover truth; they regurgitate patterns, biases and all. And when those patterns are from highly curated, propaganda-laden datasets, well, you get what you pay for.

What happens when regimes get their hands on this much data? They start building profiles. Not just for marketing, but for monitoring. For suppression. For predicting dissent before it even bubbles to the surface. We've seen it with credit scores, with social credit systems. This isn’t hypothetical. It’s happening. They're not just leveraging public data; they're demanding backdoors, building their own vast repositories through coercion and control. The West's tech giants, with their insatiable appetite for growth, often turn a blind eye, or worse, actively assist, all for a slice of that sweet, sweet market access. The money talks, always has, always will. And our so-called ethical guidelines? Mere whispers in a hurricane of cash.

Surveillance State Playbook

This isn't new, of course. Governments have always wanted more control, more eyes. But now, they have the tools we forged. They’re not just building rudimentary systems; they’re deploying sophisticated Edge Computing solutions, pushing AI inference directly to street cameras, drones, and even wearable devices. The latency is practically nil. Real-time monitoring. Real-time identification. It’s the stuff of dystopian novels, only it’s running on GPUs and server racks right now, funded by Western capital, often designed by Western engineers.

These regimes have figured out the playbook: collect everything, analyze it with AI, then act decisively.

• Facial Recognition at Scale: Not just for airports, but for every street corner. Linking faces to national IDs, social credit scores, familial networks.
• Sentiment Analysis on Steroids: Monitoring social media, encrypted chats (where possible), forum posts. Flagging dissent, identifying influencers, mapping opposition networks. LLM Hallucinations? Not an issue when the goal isn't truth, but identifying potential troublemakers.
• Predictive Policing 2.0: Moving beyond simple hotspots. Using historical data, behavioral patterns, and even biometric inputs to identify individuals "at risk" of committing crimes or, more nefariously, "thought crimes."
• Digital Censorship Ecosystems: AI-powered firewalls that adapt in real-time, blocking not just keywords but entire conceptual narratives. Dynamic content filtering. They're constantly evolving, making traditional circumvention tactics obsolete.

The tech industry's response? A collective shrug and a mumbled apology about "unintended consequences." Unintended? We designed these systems to be powerful, to be efficient. We knew the double-edged sword was sharp. We just chose to ignore the potential for harm because the CAPEX was high and the ARPU looked promising. We drank the Kool-Aid, believing that technology was inherently neutral, a force for good. Turns out, it's just a force. And forces can be wielded by anyone with enough will and lack of conscience. These despots? They have both in spades.

AI-Washing the Atrocities

Here's the rub: they’re not just using AI for surveillance; they’re using it to legitimize their actions. When an algorithm flags someone, it carries a veneer of scientific impartiality. "The system identified X as a threat." Suddenly, arbitrary arrests become data-driven decisions. Mass detentions become a "necessary security measure" backed by sophisticated analytics. It’s a convenient shield, a technological smokescreen for human rights abuses. We, the tech community, provided the tools for this optical illusion.

The PR spin from these nations is masterful. They trumpet their "innovative" use of AI to "maintain stability," to "ensure social harmony." They talk about their "responsible AI frameworks" that are often just thinly veiled justifications for authoritarian control. And guess who provides the academic legitimacy? Western universities, often through research partnerships, sometimes unknowingly, sometimes for a quick buck, always with a detached air of intellectual curiosity, ignoring the blood on the hands that fund their labs. Total nonsense. But we buy it anyway.

It's not just about what AI does, but what it enables. It accelerates oppression. It scales injustice. It automates prejudice. And it does it with a speed and a scope that human bureaucracies could only dream of. The old guard despots, they had to rely on legions of informants and cumbersome files. The new guard? They have AI, churning through billions of data points a second, flagging, categorizing, targeting. It makes their job terrifyingly efficient. And we, in our hubris, built them the perfect machine.

Engineered Reality, Engineered Dissent

The endgame isn’t just control; it’s the active construction of an alternative reality. With generative AI, with deepfakes, with highly sophisticated content filters, these regimes can manipulate public perception on an unprecedented scale. They don’t just block dissent; they preempt it by flooding the information ecosystem with state-approved narratives, personalized propaganda, and synthetic realities that reinforce their power. The truth becomes a moving target, an illusion generated by algorithms. This isn't just censorship; it's reality engineering.

• AI-Generated Propaganda: Create highly realistic, emotionally resonant articles, videos, and social media posts, tailored to individual psychological profiles. Mass persuasion.
• Deepfake Weaponization: Discredit opposition figures with fabricated videos, spread disinformation, sow distrust in authentic media. Erosion of trust in anything visual or auditory.
• Algorithmic Echo Chambers: Craft personalized information bubbles where citizens only see what the state wants them to see, reinforcing their narrative and isolating them from dissenting views.
• Automated Narrative Control: Deploy bots and AI agents to engage with citizens online, shape discussions, and counter opposition narratives in real-time.

We've created tools that allow for an unprecedented level of informational control. And the people who are best at wielding them are not the benevolent giants of Silicon Valley, but the ruthless pragmatists in authoritarian capitals. We built the perfect digital panopticon, and then we handed the blueprints to the jailers. Our idealism about open source and free information? That ship sailed a long time ago. Now, it’s about control, plain and simple. We should have seen this coming. Many of us did. But the lure of innovation, the promise of the next big thing, it blinded us. And now, we’re left watching the fallout, wondering what kind of world we’ve actually wrought.

FAQ: The Hard Questions

"But surely Western democracies have safeguards against this kind of AI misuse?"

The Blunt Truth: Safeguards? We have suggestions. And committees. While democratic nations debate ethics, the despots are deploying. Our regulatory frameworks are always playing catch-up, always reactive. They're built on the assumption of good faith, a luxury not afforded in the real world.

• Quick Fact: Many Western-developed AI tools, especially for surveillance, are sold to authoritarian regimes by private contractors.
• Red Flag: The speed of tech innovation far outstrips the speed of legislative action.
• Quick Fact: Political will to curb surveillance tech often crumbles in the face of perceived "national security" threats.

"Can't ethical AI researchers embed safeguards directly into the models?"

The Blunt Truth: You can try. But with open-source models and the ability to fine-tune, any "safeguard" can be reverse-engineered or simply removed. The code is out there. They'll just strip out your good intentions if it serves their purpose. It's an arms race where the moral side starts with a massive handicap.

• Quick Fact: Large language models are highly adaptable; fine-tuning can quickly change their behavior or ethical constraints.
• Red Flag: The concept of "ethical AI" often becomes a marketing buzzword rather than a deeply ingrained design principle.
• Quick Fact: Adversarial attacks can often bypass supposed safety mechanisms in AI systems.

"Is there any way to roll back these capabilities or prevent further misuse?"

The Blunt Truth: Roll back? The genie's out of the bottle, sealed with open-source licenses and global adoption. Prevention? Short of a global, enforced moratorium on certain types of AI research and deployment – which will never happen – we’re essentially playing whack-a-mole. The technology isn't going away. The best we can hope for is to empower the resistance, make the tools for freedom as powerful as the tools for oppression. A long shot, I know.

• Quick Fact: Key AI technologies, once developed, are incredibly difficult to contain or restrict globally.
• Red Flag: The profit motive ensures continuous development, regardless of ethical concerns.
• Quick Fact: True international cooperation on AI regulation is currently a pipe dream due to geopolitical tensions.

Parting Shot: In the next five years, expect the line between digital surveillance and physical control to blur entirely. Expect AI-powered propaganda to be indistinguishable from reality, making objective truth a relic. And expect the architects of this brave new world, both the despots and the Silicon Valley titans who sold them the rope, to continue reaping immense rewards while the rest of us grapple with the chains. We built a beautiful cage, didn't we? And now, the wardens are here.