Autonomous AI agent finds four EPICS vulnerabilities, including critical RCE flaw

An autonomous AI agent has discovered four vulnerabilities in the EPICS open source software toolkit, including a critical flaw that could allow remote code execution on systems used to control particle accelerators and fusion experiments. All four affect EPICS Base, and organizations should update to version 7.0.11, which includes the fixes.

Critical and high severity flaws

The most serious issue is CVE-2026-114595, a critical integer overflow rated 9.8 under CVSS v3.1. It can lead to an attacker-controlled out-of-bounds heap write and potentially remote code execution. The other three flaws are rated high severity. CVE-2026-114596 is an 8.2-rated improper message-length validation flaw that causes an out-of-bounds memory write and a controller process crash. CVE-2026-114597 is a 7.5-rated insufficient bounds checking bug that allows out-of-bounds memory reads and process crashes. CVE-2026-114598 is a 7.5-rated type confusion flaw that can lead to denial of service and could expose adjacent memory contents.

The US Cybersecurity and Infrastructure Security Agency issued a Tuesday advisory about the EPICS vulnerabilities. CISA warned that successful exploitation could allow an attacker to cause a denial-of-service condition, corrupt memory, disclose sensitive information or potentially execute arbitrary code.

Agent found bugs in under four hours

Causal Security co-founder Shubham Raj said an AI agent called Cipher independently found all four previously unknown security holes. The agent investigated and tested the unfamiliar EPICS system and validated the four exploits in three hours and 49 minutes. Raj said the initial investigation required no human intervention, followed by one instruction authorizing deeper exploitability testing in a secure environment.

Cipher was previously used to find multiple critical attack paths in power-grid systems used across North America and Europe. That work prompted a CISA security advisory on October 8. Raj described Cipher as a fully autonomous super hacker that can reason across the technology stack.

Exposure and update concerns

EPICS was originally developed by Los Alamos National Laboratory and Argonne National Laboratory. It is used to build and remotely control distributed SCADA systems that connect operators to physical devices. The toolkit is used in around 90 major scientific, university and industrial projects and facilities globally, including Argonne National Laboratory's Advanced Photon Source and ITER's fusion reactor control system.

Raj said the prerequisite for exploiting these vulnerabilities at a real scientific facility is being able to reach the vulnerable EPICS services on a control network. This could occur if networks were not sufficiently segmented or if another system had access to the vulnerable EPICS network. He added that exploitation would not require deep operational technology knowledge or an understanding of how particle accelerators or fusion reactors work. Raj noted that someone with conventional exploitation skills could investigate the flaws, and increasingly capable AI systems could help develop exploits without specialized OT expertise.

According to Raj, an attacker does not need to understand the physics of a particle accelerator to disrupt the software controlling it. Crashing critical processes or interrupting monitoring and control can already have serious operational consequences. The flaws also show how AI agents are effective at finding vulnerabilities in industrial and operational technology systems that once relied on obscurity for protection.

Sources: The Register