The Indexer.deserialize_from function in retrieval_lm/src/index.py is vulnerable to deserialization of untrusted data when processing the index_meta.faiss argument. Attackers can supply crafted data that is deserialized without validation, potentially leading to arbitrary code execution or other unauthorized actions if malicious serialized objects are loaded. The CVE description states the exploit may be launched remotely and is publicly disclosed.

This vulnerability affects the AkariAsai self‑rag project, specifically the retrieval_lm component. Any installation based on a commit before 1fcdc420e48f50a7d7ab1ece5494221b93252e99 is considered vulnerable. The project follows a rolling release model and therefore no fixed or affected versions are listed.

The CVSS score of 5.3 indicates moderate severity, while the EPSS score of less than 1% suggests a very low probability of exploitation in the wild. The vulnerability is not listed in CISA’s KEV catalog, which further reduces its current risk. Attackers can launch the exploit remotely by manipulating the index_meta.faiss file. Since the issue is publicly disclosed and remains unpatched, the risk remains moderate pending a fix. The lack of an official patch or workaround means administrators should treat the vulnerability as a potential threat to systems that may load arbitrary index_meta.faiss files.