Speaker discloses GPG vulnerabilities, notes some fixed and others remain unpatched after 39c3 talk

The speaker disclosed a set of GPG vulnerabilities discovered in 2025, presenting them at the 39c3 conference. Some of the reported bugs, like memory corruption in the basic PGP message parser, received patches from the maintainers. However, not all issues were addressed in code; certain flaws remained open after disclosure. This split between fixed and unfixed issues changed the perceived reliability of the GPG toolchain for users who depend on it for cryptographic operations.

Adopting GPG now carries additional verification costs: engineers must review which specific vulnerabilities have been mitigated and which persist. The unpatched signature-spoofing flaw, for example, can be exploited when a naïve GPG tool validates a signature, meaning that trust in signed messages cannot be assumed without extra checks. Organizations may need to layer additional validation steps or consider alternative implementations to mitigate the risk introduced by the unresolved bugs.

Where GPG stops working is evident in the demonstration of footguns that the maintainer refuses to fix with code, instead labeling the associated feature as harmful in a blog post. This approach leaves users exposed to attacks that rely on those features, especially in workflows that invoke GPG directly without wrappers. The persistence of such issues means that any system assuming GPG’s correctness for signature verification may fail silently when confronted with crafted inputs.

The talk also places these events in the broader state of security for 2026, touching on responsible disclosure practices and the role of AI/LLMs in identifying similar bugs. It suggests that while researchers are not doomed, the response from upstream projects can significantly affect the effectiveness of disclosure. Engineers should weigh both the technical patches and the project’s communication style when deciding whether to rely on GPG for critical security functions.