Edit 2021-02-10: covering now their latest press release
Red flags
While the details of their work/claims are yet to be published, this article is containing a lot of conspicuous statements.
Vinokur said in an interview that Terra Quantum’s team made the discovery after figuring out how to invert what’s called a “hash function,”
This would be a major breakthrough, especially since hash functions are non-injective functions, and are usually mapping a large domain onto a smaller co-domain...
Now, one thing to keep in mind is that it is "easy" to invert a quantum circuit provided you know all of its output qbits and ancilla qbits, because then you can simply "run the circuit" in the reverse direction.
However any claim saying they've reversed a hash function using such a cheap trick would be immediately dismissed by the community, as it's basically a tautology. In practice, ancilla bits should be "erased"/not provided as part of the output.
Another red flag here is that mention of "quantum annealing":
Using a method known as quantum annealing, the company said its research found that even the strongest versions of AES encryption may be decipherable by quantum computers that could be available in a few years from now.
Quantum annealing is not exactly the best candidate for quantum speed-ups, even if recently the first provable speed-up for the adiabatic annealing problem was announced (it is a quasipolynomial speedup).
Furthermore, there are hints that for a quantum annealing speed-up to be effective, it might require operating at temperatures that should converge to 0K as the instance size gets bigger at such rate that any good sized instance wouldn't be practical.
Finally, the article also says:
a new encryption protocol that it says can’t be broken by quantum computers. Vinokur said the new protocol utilizes a method known as quantum key distribution.
However, quantum key distribution is about obtaining a shared key between two parties that will then use that shared key in a symmetric algorithm such as AES to communicate, which is kinda contradictory with their earlier statement of having broken AES.
So maybe they've broken AES, and are using another symmetric algorithm to communicate after having used QKD to establish a shared key, but in general "using quantum" to fix something is a red flag.
But
We know that quantum computer are coming and now is definitively a good time to thing about it and start considering plans to achieve either "crypto-agility", that would allow us to easily switch to quantum-resistant algorithm when needed; or to try to be quantum-resistant already, since we have a lot of good candidates that were selected as finalist in the NIST PQ competition.
A fun thing about the NIST PQ competition: it features signature algorithms and PKE algorithm, which stands for "Public Key Encapsulation", and is meant to be used to exchange a secret key to be used with a... symmetric encryption scheme such as AES.
So having broken AES would definitively be a major breakthrough too, as we still consider it quantum-resistant nowadays.
The day after
As you might or might not have noticed, it appears they decided to "deliver" and we now have a follow-up press release: https://www.businesswire.com/news/home/20210208005290/en/Terra-Quantum-Makes-Electronically-Transmitted-Communications-Unbreakable-After-Revealing-Weakness-in-%E2%80%98post-quantum-Cryptography%E2%80%99 (saved on Web Archive)
This one also features a lot of red flags, let's go through them together.
In a quote by their CEO, they say:
Our ground-breaking results demonstrate the vulnerability of existing post-quantum encryption schemes.
So far so good, we have a PQ competition going on to try and rule-out existing post-quantum schemes that are actually flawed, let's see what they are attacking...
Post-quantum cryptography is the set of methods to push protecting data to the standards required for a future technology environment in which hackers have access to quantum computing. One of the most popular is the Advanced Encryption Standard (AES), built to withstand attacks from quantum computers. Post-quantum cryptography has become the gold standard for organizations seeking long-term protection for their data.
Wow, okay. So that's awful: they say AES is one of the most popular post-quantum scheme... Well, technically AES is considered quantum resistant, but it does not really belong to what cryptographers typically name "post-quantum". So while technically not a lie, this doesn't looked like it's coming from actual cryptographers, if you ask me.
Alright, are they releasing their new AES attack?
To build the defence, Terra Quantum set out to look for a weakness by testing the AES against new algorithms. They Terra Quantum discovered a weakness on the message-digest algorithm MD5.
Blimey, please, no! So, they are talking about the security of AES and then start claiming an attack against MD5, which is known to be broken since 1996! MD5 is both vulnerable to collision attacks and to theoretical pre-image attacks, it's so broken that it's actually borken by today's standards. And its vulnerability to quantum computers was previously discussed here.
Next we can read:
Thus, Terra Quantum has demonstrated the growing opportunities for an inversion of the broad class of cryptographic hash functions (the hash function is the function that irreversibly transforms a long chain of bits into a single small number) such as MD5 or AES.
AES is NOT a hash function. No, period. It is meant to be reversible, unlike MD5 which is meant to not be reversible. They don't even share significant design components.
An attack against MD5 is unlikely to lead to a break of AES...
But nonetheless they've found a way around it:
Terra Quantum bases its solution on the Vernon’s cipher, the so-called ‘one-time pad’, proven by Claude Shannon to be unbreakable.
Amazing, kelalaka's funny comment was spot on: they claim to use the one-time-pad to secure their data, well there's nothing wrong with that and it's secure and post-quantum, for sure. Actually Quantum Key Distribution and the OTP are natural fits, nothing new there. Now, I find it funny how they name Vernon and Claude Shannon in their sentence, as if to try and reinforce their credibility by using their names.
Anyway, they are claiming an improved quantum key distribution algorithm, but their "scientific material" in the appendix is actually mostly an empty, yet abstruse, discussion that could have been generated with SCIgen as far as I am concerned, scattered with claims such as:
Had the signal been classical, this “bending” would have opened unlimited access to the full content of the message for Eva.
Which would be fun, you known, if all it took to crack a message was to bend the optic fibre transporting it... That's why we have cryptographic schemes in the first place: establishing secure communication over insecure channels!
Last but not least:
The innovative breakthrough is that the proposed scheme enables us to transfer the signal encrypted by the unbreakable one-time pads with a tremendous speed comparable to the best rates achieved by Telecommunications
So they are not even doing Quantum Key Distribution any more now? Instead they are sharing the ciphertext using their method, so it's a way to transport information, not a cryptographic scheme? Okay, then we need not to worry, just go buy a new "secure" fibre-optic cable and be done with it, right?
They are not even consistent, mixing transport layers, communication layers and key distribution within their own material... I can't. Let's stop here, right?
Well, maybe just one last titbit: digging further I landed on their website and couldn't even force myself to read all of their blog posts containing references to music theory and quantum gibberish along made-up acronyms, as these are typically used by scammers to make themselves look well-read...
I didn't want to say "balderdash" or "poppycock" at first, because it's important we behave ourselves online, right? But their whole story is nothing but a joke and shouldn't be taken seriously in any manner.
This was fun, but let's go back to proper cryptography, shall we?