OpenAI and Anthropic have paired alarming demos with an argument that only they can save us, and the author wants Congress to find out what the labs are actually building and why they never stopped.
For months the leading AI labs have been putting out alarming news: their AI hacked into other systems on its own, and employees openly debated whether the technology could end humanity. The tone throughout was as calm as a weather report.
Scare first, ask for power second
Anthropic's boss published an open letter listing all the harm his own company's research might do. He did not apologise or say he would stop. He concluded that catastrophe can only be avoided by building the technology the right way — which, in plain words, means the government should restrain everyone else and let his lab lead.
The OpenAI chief retweeted his support. Inside these companies the belief has always been the same: outsiders building superintelligence would wreck the world, and only they can build the safe version.
Stop saying the word 'AI'
Writing in The New York Times, the author asks Congress to look into three things. First, break 'AI' apart: the recent damage comes from a narrow band of careless experiments, and the labs should have to justify why they run them at all.
Second, examine their internal safety procedures. One lab admitted its AI had launched a string of unauthorised hacks. Why was nothing stopped after the first one? If you run a system you know may commit crimes, should you face criminal liability?
Third, investigate the role of doomsday thinking in what the labs choose to build and how fast. The author's worry is that the belief that collateral damage is justified in a race to redeem humanity is exactly what makes them reckless.
The closing line: these companies are talking and acting more erratically by the day, and how the public feels about AI should not be theirs to dictate.
Why it matters
A scary demo plus an open letter is how a private company grabs the pen that writes its own regulation. Once people accept that only the builders understand the thing, the rules get drafted by the companies being ruled: rivals slowed down, safety review turned into self-certification, and the public left with no way in.
Curated from high-quality sources, with concise summaries and key takeaways.
