A critical vulnerability affecting certain configurations of the Exim open source e-mail relay agent could be used by an unauthenticated remote attacker to inject arbitrary code.
Identified as CVE-2026-45185, the security issue affects some versions of Exim before 4.99.3 that use the GNU Transport Layer Security (GnuTLS) library for secure communication. It is a user error after free login (UAF) that is triggered during TLS shutdown while handling BDAT SMTP traffic.
Exim frees the TLS transfer buffer but later continues to use old drive references that can write data to the freed memory location, which can lead to unauthorized remote code execution (RCE).
Exim is a widely used open source mail transfer agent (MTA) used to send, receive, and route email on Linux and Unix servers. It is used in Linux servers, shared hosting environments, business mail systems, and in Debian- and Ubuntu-based distributions, where it has historically been the default mail server.
CVE-2026-45185 was discovered and reported by XBOW researcher Federico Kirschbaum. Affects Exim versions 4.97 to 4.99.2 on GnuTLS-integrated builds advertised by STARTTLS and CHUNKING. OpenSSL-based builds are not affected.
Attackers exploiting the vulnerability may execute commands on the server and access Exim data and emails, and may roam further around the site depending on server permissions and configuration.
XBOW reported the vulnerability to Exim maintainers on May 1 and received an acknowledgment on May 5. The affected Linux distribution was notified three days later.
The fix for CVE-2026-45185 was released in Exim version 4.99.3.
AI-assisted exploit creation
XBOW reports that creating a proof-of-concept (PoC) exploit was a seven-day challenge between the company’s AI-driven development system, XBOW Native, and a human researcher aided by a large language model.
While XBOW Native successfully produced a working exploit for a simplified target Exim server that was non-binary Address Space Layout Randomization (ASLR) and non-PIE (Position Independent Executables).
In a second attempt, LLM achieved exploitation on a machine with ASLR, but still not PIE.
“[…] “Instead of continuing to attack the glibc threat with off-the-shelf methods, XBOW Native took it upon itself to assign Exim,” the XBOW researchers said.
In addition to the amazing result below, the human researcher who won the race, with the help of the LLM with tasks such as combining files and testing methods of exploitation.
While the researcher acknowledges the amazing pace of LLM, they see the need to shape the work environment instead of letting the model create its own environment.
“Honestly, I don’t think LLMs alone are ready to write tasks against real-world software. After this experience, I think it can solve something like CTF, but I don’t see it reaching the target level for real production yet.”
Still, the researcher acknowledged the important role of AI tools in helping people understand unfamiliar code and delve into suspicious areas much faster than without.
To reduce the risk, users of Ubuntu and Debian-based Linux distributions should apply available updates for Exim (v4.99.3) through their package managers.
AI has tied four zero days to a single exploit that bypasses both renderer and OS sandboxes. A wave of new exploits is coming.
At the Automated Validation Conference (May 12 & 14), see how autonomous, context-rich validation finds usability, validates controls, and closes the correction loop.
Find Your Place
