Skip to content

Security: MuteJester/AlignAIR

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please report security issues privately to thomaskon90@gmail.com rather than opening a public issue. Include a description, reproduction steps, and the affected version. We aim to acknowledge reports within a few business days.

Scope

AlignAIR distributes models as .alignair files. The sections used for inference — the config (JSON), the weights (safetensors), and the embedded germline reference (reference_json) — are portable and load without executing any pickle, and the embedded reference's allele-order and FASTA fingerprints are hash-verified on every load. Models published for inference contain zero pickle sections.

A .alignair file may additionally carry resumable-checkpoint sections (dataconfig, train_state) that use Python pickle / torch.load and therefore execute code on load. Only load checkpoints — or any model file — from sources you trust. Models downloaded from the official catalog are fetched over HTTPS from the Hugging Face Hub and verified against the catalog's SHA-256.

There aren't any published security advisories