Baseband firmware/processors are still making strides in terms of defending against attacks so I don't think we are at the point where backdoors are needed. More likely that undocumented AT commands, plausibly deniable bugdoors and existing memory corruption vulnerabilities are already convenient reality. However, part of that reality is also that once the cellular modem is exploited OTA it is still a massive task to traverse IOMMU restrictions and get device owner data/secrets from the userspace OS. It doesn't even grant visibility into application traffic which has been aggressively TLS encrypted since the mid 2010s.
The entities that are putting in backdoors don't respect the law, and obviously will do what they can to obfuscate the backdoor enough to pass regulatory reviews.
My threat model is precisely entities that don't respect law or are unethical. So, the question then boils down to whether current regulatory review processes can catch such backdoors and the people running these processes cannot be bribed or otherwise forced to skip proper review?