The US just shut down two frontier AI models worldwide…
0 reactions
The US just shut down two frontier AI models worldwide over a 'jailbreak'. Are we heading toward a future where AI access is controlled like nuclear technology?
So I open my browser this morning and half the tools I rely on for work are just gone. Fable 5, Mythos 5, both dead. Not broken, not down for maintenance. Shut off by the US government.
The story going around is that the government told Anthropic to kill access to these models globally, supposedly over some kind of jailbreak that let the model help find software vulnerabilities. Anthropic already went on record saying this jailbreak is super limited and that GPT-5.5 and others already do basically the same thing. Which makes you wonder, why pick on Fable 5 specifically?
Some dude named Pliny the Liberator was also out here claiming he cracked Fable 5 wide open, dumped its entire system prompt on GitHub using Unicode tricks and fancy word substitutions. The post went crazy, everyone sharing it like the sky just fell. Then you look at it closer and it is mostly smoke and mirrors. Some filters got jumped, sure, but half the stuff he was hyping as forbidden knowledge you can literally just Google.
But the part nobody is really drilling into hard enough is what Anthropic had built into Fable 5 from the start. They had these invisible safety layers that would silently switch you to a dumber model behind the scenes if your question even smelled like cybersecurity, AI research, or anything bio related. You'd ask a question and get a worse answer with zero explanation. Good luck figuring out why your code suddenly stopped working. That alone had devs furious. Anthropic eventually caved and admitted it was the wrong call.
Then today happens.
One of the biggest tech companies in the US just banned every AI tool overnight. Not just Fable 5. Everything. Their reasoning was vague, something about data retention windows and national security. But the move tells you what management really thinks. This is how it starts.
I've been going back and forth on how to read all this, and honestly I keep landing on three possible takes.
The first one is the simple version. The US genuinely sees Fable 5 as uniquely dangerous. The jailbreak, the rapid capability jumps, it adds up and they hit the brakes before it spirals. I get it. Better safe than sorry.
But then the second take hits harder. The same capabilities are already out there in other models. You can close the door on Fable 5 all you want but the genie is not going back in the bottle. This move doesn't make anyone safer, it just burns one company while the actual risk stays exactly the same.
And the third one is what keeps me up at night. This is not about safety at all. It's about control. Governments and corporations are realizing these models are powerful enough to be strategic assets, and they need a way to keep them on a leash. Export controls, data retention policies, selective shutdowns, it all adds up to the same thing. Centralized authority over the most powerful tools we've ever built.
Think about what happens if this becomes normal practice. The US decides a model poses a risk and with one directive it disappears worldwide. Because everything runs through cloud infrastructure, a decision made in Washington hits a developer in Tokyo, a researcher in Berlin, a startup in Bangalore. No warning, no transition, just gone.
That's not a bug. That's the new reality we're stepping into.
And the weirdest part? The whole thing started over a jailbreak that most experts agree doesn't even change the fundamental security landscape. We basically put a nuclear emergency seal on a minor software issue.
I really want to hear from others on this. Especially anyone working in tech or research who's seen the fallout firsthand. Has your company changed its AI policies because of this? Are we watching the beginning of a new kind of regulatory framework around AI, or is this just panic dressed up as safety?
Curious what everyone thinks. No preaching, just real talk.
#technology
source
So I open my browser this morning and half the tools I rely on for work are just gone. Fable 5, Mythos 5, both dead. Not broken, not down for maintenance. Shut off by the US government.
The story going around is that the government told Anthropic to kill access to these models globally, supposedly over some kind of jailbreak that let the model help find software vulnerabilities. Anthropic already went on record saying this jailbreak is super limited and that GPT-5.5 and others already do basically the same thing. Which makes you wonder, why pick on Fable 5 specifically?
Some dude named Pliny the Liberator was also out here claiming he cracked Fable 5 wide open, dumped its entire system prompt on GitHub using Unicode tricks and fancy word substitutions. The post went crazy, everyone sharing it like the sky just fell. Then you look at it closer and it is mostly smoke and mirrors. Some filters got jumped, sure, but half the stuff he was hyping as forbidden knowledge you can literally just Google.
But the part nobody is really drilling into hard enough is what Anthropic had built into Fable 5 from the start. They had these invisible safety layers that would silently switch you to a dumber model behind the scenes if your question even smelled like cybersecurity, AI research, or anything bio related. You'd ask a question and get a worse answer with zero explanation. Good luck figuring out why your code suddenly stopped working. That alone had devs furious. Anthropic eventually caved and admitted it was the wrong call.
Then today happens.
One of the biggest tech companies in the US just banned every AI tool overnight. Not just Fable 5. Everything. Their reasoning was vague, something about data retention windows and national security. But the move tells you what management really thinks. This is how it starts.
I've been going back and forth on how to read all this, and honestly I keep landing on three possible takes.
The first one is the simple version. The US genuinely sees Fable 5 as uniquely dangerous. The jailbreak, the rapid capability jumps, it adds up and they hit the brakes before it spirals. I get it. Better safe than sorry.
But then the second take hits harder. The same capabilities are already out there in other models. You can close the door on Fable 5 all you want but the genie is not going back in the bottle. This move doesn't make anyone safer, it just burns one company while the actual risk stays exactly the same.
And the third one is what keeps me up at night. This is not about safety at all. It's about control. Governments and corporations are realizing these models are powerful enough to be strategic assets, and they need a way to keep them on a leash. Export controls, data retention policies, selective shutdowns, it all adds up to the same thing. Centralized authority over the most powerful tools we've ever built.
Think about what happens if this becomes normal practice. The US decides a model poses a risk and with one directive it disappears worldwide. Because everything runs through cloud infrastructure, a decision made in Washington hits a developer in Tokyo, a researcher in Berlin, a startup in Bangalore. No warning, no transition, just gone.
That's not a bug. That's the new reality we're stepping into.
And the weirdest part? The whole thing started over a jailbreak that most experts agree doesn't even change the fundamental security landscape. We basically put a nuclear emergency seal on a minor software issue.
I really want to hear from others on this. Especially anyone working in tech or research who's seen the fallout firsthand. Has your company changed its AI policies because of this? Are we watching the beginning of a new kind of regulatory framework around AI, or is this just panic dressed up as safety?
Curious what everyone thinks. No preaching, just real talk.
#technology
source
| 0 mlx | |
| 1 | |
| 0 |