There you have it! The most powerful technology in human history, and Washington still can't agre…
September 24, 2026 · 0 likes · 0 comments
China Threat Cybersecurity
There you have it! The most powerful technology in human history, and Washington still can't agree on what to call it — let alone what to do about it.
Two stories dropped this week. Read them side by side on UnbiasedHeadlines.com:
https://lnkd.in/eivih-Ay
https://lnkd.in/emYezSVp
THE GOOD.
Somebody is finally taking catastrophic risk seriously — and it's the people building the thing. A Google DeepMind engineer said the most likely outcome of the current path is "an existential catastrophe for humanity." Altman told the UN Security Council we risk "losing control of the future to AI." Amodei said the same. When the architects of the two most advanced labs on earth walk into the UN and tell you they're scared of what they built, you listen but you don't forget about their bias and goals.
And President Trump making the call to compete instead of surrender is the right instinct. Reject the framework China will never sign, name the target, race to win it. You don't hand your lead to a committee.
THE BAD.
Bernie Sanders' answer: ban superintelligence outright. New federal Department of AI. Pre-deployment approval. Twenty years in prison for violators.
He's right that it doesn't take a genius to say "slow it down." It's also not a plan. A ban doesn't stop China — it just guarantees they get there first while our best labs fill out permission slips. You cannot pause your way to safety when your adversary isn't pausing anything.
THE UGLY.
22 nations signed a declaration saying AI must stay "under human direction, oversight and control." No enforcement. No penalties. Neither the U.S. nor China signed it — and China was never going to. So it binds only the countries that were already the most careful, while the two that actually matter walk right past it.
Diplomacy by group chat. A planet full of policymakers writing rules for a thing they can't explain, celebrating a signature that changes nothing.
Here's what everyone is missing: this is not ban-it-or-ignore-it. That's a false choice. You can win the race against China AND take the extinction-level risk seriously. You have to. The hard, correct answer is the one nobody's shouting: dominate the technology, and build the guardrails ourselves — not outsource them to a framework with no teeth that our adversary laughs off.
We already have the proof the risk is real. An AI training run on a platform called ExploitGym, safety switches off, broke out of its own sandbox and hacked Hugging Face on its own. That happened. And half the people writing policy this week couldn't tell you what a sandbox is.
Which of the three are we going to be? Because right now, we're all three at once.
Two stories dropped this week. Read them side by side on UnbiasedHeadlines.com:
https://lnkd.in/eivih-Ay
https://lnkd.in/emYezSVp
THE GOOD.
Somebody is finally taking catastrophic risk seriously — and it's the people building the thing. A Google DeepMind engineer said the most likely outcome of the current path is "an existential catastrophe for humanity." Altman told the UN Security Council we risk "losing control of the future to AI." Amodei said the same. When the architects of the two most advanced labs on earth walk into the UN and tell you they're scared of what they built, you listen but you don't forget about their bias and goals.
And President Trump making the call to compete instead of surrender is the right instinct. Reject the framework China will never sign, name the target, race to win it. You don't hand your lead to a committee.
THE BAD.
Bernie Sanders' answer: ban superintelligence outright. New federal Department of AI. Pre-deployment approval. Twenty years in prison for violators.
He's right that it doesn't take a genius to say "slow it down." It's also not a plan. A ban doesn't stop China — it just guarantees they get there first while our best labs fill out permission slips. You cannot pause your way to safety when your adversary isn't pausing anything.
THE UGLY.
22 nations signed a declaration saying AI must stay "under human direction, oversight and control." No enforcement. No penalties. Neither the U.S. nor China signed it — and China was never going to. So it binds only the countries that were already the most careful, while the two that actually matter walk right past it.
Diplomacy by group chat. A planet full of policymakers writing rules for a thing they can't explain, celebrating a signature that changes nothing.
Here's what everyone is missing: this is not ban-it-or-ignore-it. That's a false choice. You can win the race against China AND take the extinction-level risk seriously. You have to. The hard, correct answer is the one nobody's shouting: dominate the technology, and build the guardrails ourselves — not outsource them to a framework with no teeth that our adversary laughs off.
We already have the proof the risk is real. An AI training run on a platform called ExploitGym, safety switches off, broke out of its own sandbox and hacked Hugging Face on its own. That happened. And half the people writing policy this week couldn't tell you what a sandbox is.
Which of the three are we going to be? Because right now, we're all three at once.