OpenAI's Government Push Is the Most Predictable Move in Tech History
The company that wants to build AGI is now cozying up to governments and newsrooms. I've seen this movie before, and the ending's never as clean as the pitch deck.
画像クレジット: Lottie animation by Centre Robotics (LottieFiles Free, used with credit). · source
OpenAI wants to be everywhere. And I mean everywhere. Schools in Greece, newsrooms in Singapore, government offices across America. If you've been paying attention to how tech giants operate (and I have, for longer than some of these founders have been alive), none of this should surprise you. It's the platform play, dressed up in the language of public good.
Let me be clear: I'm not saying this is necessarily bad. But let's not pretend it's altruism either.
The government angle is the big one. OpenAI recently launched something called "OpenAI for Government," which is exactly what it sounds like: a push to get ChatGPT and related tools into the hands of U.S. public servants. The OpenAI blog post talks about "best-in-class technology" and "service of the public good," which is the kind of language that makes my teeth itch. I've heard it from Microsoft in the 90s, Google in the 2000s, and now OpenAI in the 2020s. The playbook doesn't change, just the product.
What does change is the stakes. We're not talking about office productivity software here. We're talking about AI systems that governments might use for, well, we don't actually know yet. That's the problem! OpenAI hasn't been particularly specific about which agencies are interested or what they'd use these tools for. Policy analysis? Citizen services? Something more sensitive? The company didn't disclose details, and call me old-fashioned, but I think that matters.
Then there's the education push. OpenAI and the Greek government announced "OpenAI for Greece," which will bring ChatGPT Edu into secondary schools. The stated goals are AI literacy, supporting local startups, and driving economic growth. All good things, in theory. But I've covered enough ed-tech initiatives to know that the gap between announcement and implementation is where most of these programs go to die.
関連記事
More in Policy
Everyone focused on the headline rules. Annex IV is where the real pain lives.
James Chen · Yesterday · 4 min
Two major AI labs announced UK partnerships within days of each other. The details are thin, but the implications are significant.
Aisha Patel · Yesterday · 5 min
The AI giant wants to reshape American manufacturing and infrastructure policy, but the details are thin and the ambitions are enormous.
James Chen · 2 days ago · 6 min
The company has released a series of frameworks for protecting young users, but the real question is whether these guidelines have teeth or if they're primarily a PR exercise.
The Greek partnership is interesting because it's one of the first times OpenAI has gone this deep with a foreign government on education. It suggests they're testing models for broader rollout. If Greece works, expect similar announcements for other countries. If it doesn't, well, we probably won't hear much about it.
The newsroom stuff is more complicated, and honestly more interesting to me. OpenAI launched something called the OpenAI Academy for News Organizations, built with the American Journalism Project and The Lenfest Institute. The pitch is training for journalists, editors, and publishers on how to use AI "responsibly." They've also been signing partnership deals, with Axios being the latest.
Now look, I'm a journalist. I use AI tools. I'm not some Luddite screaming about the machines taking our jobs (though some of them probably will, let's be honest). But there's something a little weird about the company that's been accused of training on copyrighted news content now positioning itself as a friend and teacher to newsrooms. It's like the fox offering to teach the hens about home security.
The CNA case study is instructive here. OpenAI published a piece about how the Singapore-based news outlet is "transforming its newsroom with AI," featuring their Editor-in-Chief talking about adoption and culture change. It reads like a success story, which I'm sure it partly is. But what we don't get is the messy middle: the failed experiments, the ethical debates, the reporters who pushed back. Real transformation is never as smooth as a case study suggests.
So what's actually happening here? OpenAI is doing what every platform company does once it achieves scale: it's trying to become infrastructure. The kind of thing that's so embedded in how institutions operate that switching costs become prohibitive. Get into schools, and you shape how the next generation thinks about AI (and whose AI they reach for first). Get into government, and you become part of the bureaucratic furniture. Get into newsrooms, and you influence how information itself gets produced and distributed.
This is the self-driving car hype cycle all over again, except the stakes are different. With autonomous vehicles, the worst case scenario was cars that didn't work as promised (and some tragic accidents along the way). With AI systems embedded in government and media, the failure modes are harder to see and potentially harder to reverse.
I should note that OpenAI recently announced some leadership changes, with the company acknowledging it's "grown a lot" while trying to stay focused on frontier research. That's the tension, isn't it? You can't be both a research lab pursuing AGI and a vendor selling services to governments and newsrooms. Those are different things that require different cultures, different incentives, different accountability structures. But what do I know.
The question I keep coming back to is: who's asking the hard questions? When OpenAI partners with a government, who negotiates the terms? What data gets shared? What happens when a government wants to use these tools in ways that OpenAI finds uncomfortable? Or, more likely, what happens when OpenAI finds out years later that a government client was using their tools in ways nobody anticipated?
These aren't hypothetical concerns. We've seen this with every major tech platform that's gone down the government services road. The initial partnership is all handshakes and press releases. The problems come later, quietly, and by then the contracts are signed and the dependencies are built.
I'm not saying OpenAI is uniquely bad here. They're not! They're doing what rational actors do when they have a product that lots of people want. And some of this will probably produce genuine benefits. Greek kids learning about AI is probably good. Newsrooms having better tools is probably good. Government services running more efficiently is probably good.
But "probably good" isn't the same as "definitely good," and the gap between those two things is where the interesting questions live. Questions that OpenAI, understandably, isn't super eager to dwell on in their blog posts.
Here's what I'd tell the kids building and deploying this stuff: The technology is impressive. Genuinely impressive! I've been covering tech long enough to know when something is real versus when it's vaporware, and this is real. But impressive technology doesn't automatically produce good outcomes. The implementation matters. The governance matters. The incentives matter. And right now, we're in the phase where everyone's focused on adoption and nobody's focused on what happens after.
That's usually when things get messy.
If you want to argue with me about any of this, my email's on the about page. I still prefer it to Slack, and I read everything. Even the angry ones. Especially the angry ones, actually.