GLASS is a header-only CUDA template library; only the latest release (and main) receive fixes.

Please report vulnerabilities privately through GitHub's security advisory flow: Report a vulnerability (Repository → Security → Report a vulnerability). Do not open a public issue for a security-sensitive report.

We aim to acknowledge reports within a week. Because GLASS runs entirely inside the caller's own CUDA kernels (no network, no file I/O, no privileged operations), most issues are correctness bugs rather than vulnerabilities — but memory-safety reports (out-of-bounds shared-memory access, scratch-size under-allocation in a documented-correct usage) are in scope and welcome.