What I take issue with actalis, is that they don’t just sign your private key but you actually get the private key from them. It then depends on how much you trust the issuer.
By definition, that key can no longer be considered “private”.
It is very important to emphasize that the key in this model is not "private" anymore. Thus, all the communication using this key is not secure anymore.
Private key is the one generated by hardware owned by the user and immediately secured with strong password. Ideally, private key does not leave the hardware that generated it. Thus, every device shall have its own private key.
In less restricted model, private key gets copied by the user to other hardware using media like USB stick or P2P communication model that does not use cloud services.
Be aware, that trusted Certificate Authority (CA) configuration applies to ALL certificates issued by CA. Thus, if one elects to trust "actalis" CA, then they trust ALL actalis CA users.
If the process of obtaining certificate was extremely simple, easy and did not involve identity verification steps, then bad actors can take advantage of this process and create identities that your client application will trust.
By itself the bad actor identity is of little concern to anybody, but it can have a significant impact if trusted identity is used in spam filtering, exploits of email client bugs or other hack attempts. Trusted users may be given higher access privilege at the client application level, which may be just enough for hacker to gain required access. For example, client application may be configured to trust all trusted senders with MIME attachments. An unknown trusted user sends malicious Application as file attachment. Accidental double click lunches the application without "are you sure?" prompt. Congratulations, machine is pwned.
The problem is easily mitigated by not importing root CA for easy CAs.