The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added four vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, citing evidence of active exploitation. Two of them stand out because of who runs the affected software: REDCap, the survey and database platform that universities and hospitals use to collect research data, and LiteLLM, an open-source gateway that companies put in front of their AI model providers.
The other two entries are an Apache Axis flaw that IntelFusions reported under attack earlier this week, and a 2024 bug in Synology's DiskStation Manager (DSM), the operating system on the company's network storage boxes.
CISA's notice lists each flaw by its vulnerability class only. It does not say who is exploiting them, against which targets, or since when.
Two 9.8 bugs lead a mixed list
- CVE-2026-90436, Apache Axis (CVSS 9.8). CISA classes it as unsafe reflection. The NVD record says it affects all Apache Axis 1.x versions through 1.4 and can be reached without authentication. Axis 1 is end of life, and the project does not expect to ship a fixed 1.x release.
- CVE-2026-90817, Vanderbilt University REDCap (CVSS 9.8). CISA classes it as code injection. According to the NVD record, it allows unauthenticated remote code execution on REDCap 13.3.0 and higher, although the attacker needs to know the hash of a valid public survey.
- CVE-2026-40217, BerriAI LiteLLM (CVSS 8.8). CISA classes it as an unprotected alternate channel. The NVD record describes remote code execution through the /guardrails/test_custom_code endpoint in LiteLLM releases through 2026-04-08.
- CVE-2024-10441, Synology DSM. CISA classes it as improper encoding or escaping of output. The notice publishes no severity score, and Synology's own advisory carries the affected versions and the fix.
REDCap servers already have hunters
This is not the first time REDCap has drawn serious attention. In June, Google's threat intelligence group described a China-linked cluster it tracks as UNC6508 searching for outdated REDCap installations to break into North American research institutions, a campaign IntelFusions covered at the time. Nothing in CISA's notice ties that activity to CVE-2026-90817, and the agency has not named any actor. It does show that internet-facing REDCap servers were on a target list well before this listing.
An AI gateway is a tempting door
LiteLLM's job is to route an organization's requests to AI providers, which usually means it holds the API keys for those accounts. The project has had a rough autumn: in October a poisoned LiteLLM release was caught stealing cloud and AI secrets. CISA's listing adds a confirmed in-the-wild bug to that record, and any team that stood up a LiteLLM proxy for a pilot and then forgot about it should go and find it.
Update REDCap and LiteLLM, migrate off Axis 1
- REDCap: follow Vanderbilt's guidance and move to a fixed release. The NVD record does not name one, so confirm the version with the REDCap consortium before assuming you are covered.
- LiteLLM: update to a current release and check BerriAI's advisory for the fixed version. Restrict access to the proxy's administrative and guardrail endpoints.
- Apache Axis 1.x: there will be no release. The Apache project recommends migrating to another SOAP engine such as Apache Axis2/Java, or rebuilding Axis with the upstream patch applied.
- Synology DSM: apply the DSM update named in Synology's advisory.
Federal civilian agencies must prioritize KEV entries on publicly exposed assets under Binding Operational Directive 26-04, which also sets expectations for checking whether a system was compromised before the patch went on. That second step matters for everyone else too. A listing in the KEV catalog means somebody has already used the bug, so a patch applied today closes the door without telling you who walked through it last week.
This briefing is provided by IntelFusions for informational and defensive purposes only. It is based on sources assessed to be reliable at the time of writing, and analytic judgments carry the confidence levels indicated. Indicators of compromise are defanged; re-arm them only in controlled environments. IntelFusions is not affiliated with the organizations named and makes no warranty as to completeness or accuracy.