Ethereum adjusts its course: the network's post-quantum protection is no longer tied to Poseidon2
An important infrastructure shift is brewing in the Ethereum ecosystem. The development team has decided to abandon plans to use exclusively the Poseidon2 hash function as the foundation for the post-quantum validator key registry. This is not a spontaneous move, but rather the result of deep cryptanalysis that uncovered potential weaknesses in the class of algebraic hash functions to which Poseidon2 belongs.
The problem proved serious enough to reconsider the entire architectural strategy. Instead of staking the network's security on a single card, a whole range of alternatives is now being considered. Among the favorites are the earlier version Poseidon1, as well as the time-tested universal functions BLAKE3 and the SHA family. However, as engineers emphasize, the final decision has not yet been made: consensus within the community regarding the optimal candidate has not yet been reached.
It is telling that this reconsideration is happening against the backdrop of unprecedented precautionary measures. Earlier, the Ethereum Foundation had already allocated $1 million for an open competition to find vulnerabilities specifically in Poseidon. The fact that even after such funding and auditing, trust in Poseidon2 has not become absolute speaks to the high bar for security requirements in the context of future quantum threats.
In my view, this is a very mature decision. In the race for post-quantum resilience, it is easy to fall into the trap of "novelty for novelty's sake" and choose an exotic algorithm that looks efficient on paper but has hidden algebraic structures convenient for attacks. Abandoning a single point of failure in favor of diversifying cryptographic primitives is a sign that Ethereum treats the threat of quantum computing not as a marketing story, but as a real engineering challenge. The only question is what compromise between verification speed and cryptographic strength will ultimately be found. Given the stakes, I expect the final choice to fall on hybrid schemes combining algebraic and classical hashes.