People have short memory.
Different cast every time. Different industry. Sometimes different century. But the plot is always the same: someone wants to sell something, limit customer ability to get it, and waits for you to want it more.
This week it was Anthropic.
You saw the headlines. Mythos. Too dangerous. Can’t release it. Twelve-company coalition checking the model. Hundred million dollars committed. Very serious faces.
And my first reaction, honestly, was the same as yours. Wow. That sounds serious. Quantum revolution came early. Maybe they really are being responsible. Maybe this is the one time a company is actually putting safety over profit. It happens. Probably. Somewhere.
It took me a full day to remember I’d had that exact feeling before.
Year 2019. GPT-2. OpenAI refuses to release it because it was “Too dangerous”. The press loses its mind, and OpenAI becomes the lab that builds Things So Powerful They Scare Even The People Who Built Them. A few months later they release it. It’s fine. But by then the brand is built, the nonprofit has quietly become a capped-profit company, and billions are flowing in. The “too dangerous” announcement was timed to the fundraising round, which I’m sure is a coincidence, the way it’s a coincidence every single time.
Year 2023. GPT-4. Published a report listing all the scary things it could do. Altman went to Congress and asked, with a straight face, to be regulated. Please, restrict us. We’re too powerful. We need adult supervision. During a fundraising round. Naturally.
Then the board fired him over “safety concerns” that were never explained. Five days of the biggest free PR event in tech history. He came back a hero. Microsoft deepened its investment. Nothing about the product changed. I guess we are safe now.
Two rounds. Same company. And each time the press covered it like it was the first time anyone had ever pulled this off.
The trick only works if you don’t remember it working last time.
This is not an AI thing, by the way. This is a human thing. We’ve been falling for this every single time.
De Beers, the diamond cartel, literally used human rights to destroy their competition. When blood diamonds became a scandal, they didn’t just support regulation. They championed it. Pushed for the Kimberley Process. Told consumers to refuse uncertified stones. Looked noble doing it. Won awards, probably.
What nobody talks about: before the Kimberley Process, independent miners were flooding the market with cheap diamonds and crashing De Beers’ prices. The certification system didn’t just block conflict stones. It legislated the cheap competitors out of existence. Re-consolidated the monopoly. Kept prices exactly where they needed them.
Everybody thanked them for it. Some still do.
Tesla “gave away” all its patents in 2014. Musk wrote a blog post about how the climate crisis was too important for intellectual property hoarding. The press swooned. Standing ovation. Bravery. Leadership. All the words.
The fine print: any company that used a Tesla patent surrendered the right to ever sue Tesla for IP theft. And if everyone adopted Tesla’s engineering standards, Tesla would own the ecosystem forever. But sure, it was about the planet.
Giving everything away was the most effective way to own everything. Philanthropy speedrun.
Sometimes, things can get too cynical.
In the late ’90s, Philip Morris spent hundreds of millions telling teenagers not to smoke. Researchers later proved those campaigns were designed to fail. By framing smoking as an “adults-only decision,” they made it look like rebellion. Teens exposed to those ads were more likely to start smoking. Which is either a tragic failure of marketing or a spectacular success, depending on which quarterly report you’re reading.
While everyone debated whether the campaigns worked, the real play happened somewhere else: the ads bought political goodwill and delayed real regulation for years.
The restriction WAS the product.
I’m not bringing these up because I think you don’t know them. Most of us know, when someone points it out.
The interesting part is that knowing doesn’t help. You have a short memory.
You know GPT-4 was a stunt. But do you remember how it felt to read the headline? The genuine concern. The impressed whisper. The endless wave of “experts” in the news telling you to feel alarmed. Your job is at stake, your house, your life. The AI is there and it’s coming for your kids. That sense that the country, government, someone should save us. In the form of competition killing regulation of course.
Gone. The fact stays. The feeling wipes clean. And the next announcement doesn’t target your analysis. It targets the feeling. And the feeling has already reset. This is how the human brain is working. Otherwise we all would have only one child.
I use Anthropic’s tools every day. I’m not saying Mythos isn’t capable. I’m saying that “too dangerous to release” is a press release, not a safety decision. That the hundred million dollar coalition isn’t a restriction, it’s distribution. And that the fundraising round happening in the background isn’t a coincidence, because it’s never a coincidence.
They’ll ship it pretty soon. I’m sure. It’ll be useful. It won’t end the world. And by then the myth will already be built. The funding will already be closed. Someone will have already written a LinkedIn post about responsible AI leadership. People will use it to search for the answer for all the biggest mysteries of this life. Again.
And we’ll forget. Again.