Don’t understand the details of this or how exploitable it is in the wild
PACMAN seems to require a hardware side channel on the speculative execution of the SoC (doesn't mean at the moment that hardware access is required, though), and obviously a pre-existing software bug (which PAC would normally protect against), so it probably won't be an immediate risk for most users.
Researchers behind PACMAN also state that this seems to affect most implementations of ARM pointer authentication (introduced in ARM 8.3), PAC being one implementation. Keep in mind PAC is also just one mitigation among others, and it's not a huge surprise to see bypass attempts.
Obviously this doesn't affect Pixel devices as even the latest Tensor chip is based on ARM 8.2, and Android on arm64 takes another approach to achieve a similar goal as PAC (fine-grained CFI and shadow stacks), perhaps in a better way in some aspects. Software mitigations could be possible but they will likely impact performance so Apple's stance remains to be seen. Unless proven otherwise, M2 should have the same flaw.
The paper will be published in a few days so we'll see in details how PACMAN works. (here)