|
|
Log in / Subscribe / Register

Unsafe (t)rust?

Unsafe (t)rust?

Posted Apr 17, 2026 19:02 UTC (Fri) by Heretic_Blacksheep (guest, #169992)
In reply to: Unsafe (t)rust? by tux3
Parent article: A more efficient implementation of Shor's algorithm

The attack you link to suggests a more subtle takes on the problems with trust*, namely in transferring trust from the scientists involved into the tools used to prove the existence of such a circuit, rather than revealing the circuit's actual physical structure. It's worth a read for the philosophical points even if people can't quite follow the details of quantum computing, what they did, and why the trust angles matter.

I think Google may ultimately be naive in believing no one else is going to stumble onto the physical design, as there are intelligence agencies that even more highly resourced than they are. Once they have proof such a circuit exists, they'll be looking for it. Hell, the current administration could just send Google an NSL and demand disclosure. Google is almost certain to fold on a legal demand. Likewise, any Chinese agents already embedded in Google will know what to go looking for. Sometimes all the world needs is to have a hole pointed out, once it's pointed out you can often deduce the shape and functionality by where the interfaces connect up to the rest of the outside world. Right now, no criminal gang has the resources to go after the crypto-markets in this way. I think most people wouldn't care or even cheer if they did. Google was severely tone deaf and narrow minded in their press release in worrying about cryptocurrency. But Shor's Algorithm current unsolvability affects more than vapor currency. The world isn't ready for a sudden break in pre-quantum cryptography's usability. Most systems haven't moved to post-quantum algorithms. Previously stored secrets, even if they are safely re-encrypted later, will become vulnerable if they haven't been wiped from all storage everywhere. The destruction of all previously encrypted stored data is impossible to prove, so those that created that data would have to depend on the complete obselesence of said data. That works for data with an enforced or natural expiration date like certain kinds of credentials, but it will never be true of biometrics, health data, DNA, or any other information with indefinite shelf life.

*Interestingly, I'd already (tried to - I admit I didn't fully understand the attack itself) read the attack article from a link by another aggregator site and was going to point it out for further reading on the philosophy of nkVMs in more practical senses aside from the details of the attack.


to post comments

Unsafe (t)rust?

Posted Apr 17, 2026 19:48 UTC (Fri) by mathstuf (subscriber, #69389) [Link]

I think there's a line where such a zero-knowledge proof makes sense: warning of impending risks without being drowned out by those calling you chicken little in the meantime. It also means that if someone can improve on it rapidly (e.g., someone at the NSA or nascent PhD research can trivially shave another factor of 10 off by some "new" trick), they're at least slowed down by having to rediscover the thing. I'd feel better with a "we'll release the actual circuit in N months time, or earlier if re-discovered" disclaimer attached to it to put more weight behind the "hey, we've got a problem on the horizon" angle.

Unsafe (t)rust?

Posted Apr 17, 2026 21:38 UTC (Fri) by roc (subscriber, #30627) [Link]

> any Chinese agents already embedded in Google will know what to go looking for.

FWIW, I work for Google, and if I were to go hunting for arbitrary data I shouldn't have access to (this quantum circuit, or users' private GMail, or whatever), I'm extremely confident I would either completely fail or be caught. Google's insider threat protection is impressive. Of course if the spy is on the right team --- e.g. the team that developed the circuit --- then things are different, but that's a much narrower threat.

I don't think NSLs can be used to demand information like this circuit. See https://support.google.com/transparencyreport/answer/9713...


Copyright © 2026, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds