It depends. The ideal hardware wallet is always off, disconnected, the seed phrase is encoded, split into pieces and buried in a WWII bunker. It would be incredibly secure and unusable.

Let's take a look at U•HODL:

Fingerprint scanner (even the best one) is less secure than a long PIN code (and few people use them), but scanning only takes a moment while entering a PIN code requires tens of seconds. And after several failed scans, the device is locked.

Is Bluetooth less secure than a cable? Even if all internal communication is heavily encrypted there is still a small chance that it can be hijacked. On the other hand, the device has a MiTM protection and wireless communication is much more usable on the go.

Is it possible to get access to your data after stealing a key? It's also plausible and you may find information about attacks on existing wallets. But all vendors try to fix these vulnerabilities once they're found.

U•HODL is no less secure than any other hardware wallet, it's just designed with a user in mind.

The article devoted to design challenge and clearly states that it was almost a year before design research started, while other researches were conducted, regarding the security model principally.

Later there would be separate articles for aspects of security approaches implied within the design of U-HODL, when whitepapers are polished, finished and ready for publishing. Anyway, I would gladly answer any direct questions about the issues you might be interested in, I just don't have a whole article at the moment for your consideration.

In short, the statement is as follows: the product would be able to comply with payment processing protocols, like every banking card for example.

Additionally, as stated in the article and in the author's comment above - the product designed with the user in mind, what is no less important for customers security according to our vision of the problem.

Security is more about a process, and just finally as a result. And hardware wallet is one of the most important instruments that should help customer to build and maintain his own security protocols and behaviour patterns. If the instrument is "cutting your own hands" than it is much less likely that user would apply it most securely and regularly. The simplest example with PIN-code which needed to be entered repeatedly at every simple step or recurring operation, it is most likely that user wouldn't use device at all, or just use a shorter PIN, than doing his replay attack without measure every day. Thus we provide the convenience for user to authorize himself biometrically much faster within a reasonable timeframe, value cap and proximity limitations (very similar to major smartphone industry biometrical practice). Doing our "user experience" homework, we see our contribution to customers security and assets safety much bigger than just provide yet another "the most secure wallet". It is the most secure both from the technical and usability point of view.