Check Point has disclosed three now-patched flaws in LangGraph, the popular LangChain framework for building AI agents, that can be chained for remote code execution on self-hosted servers. The chain combines an SQL injection (CVE-2025-67644) with an unsafe msgpack deserialization bug (CVE-2026-28277): an attacker who can reach the agent's stored-state endpoint plants a malicious checkpoint that runs code when loaded. A compromised LangGraph server exposes everything the agent can touch, including model API keys, customer data, and internal network access. It is only exploitable in self-hosted deployments using the SQLite or Redis checkpointer; LangChain's managed LangSmith platform is not affected.
The ShinyHunters data-theft wave against Oracle PeopleSoft, covered yesterday, now has a confirmed root cause: a zero-day. Oracle has issued an out-of-band emergency mitigation for CVE-2026-35273, a critical flaw (rated 9.8) in PeopleSoft PeopleTools that lets an unauthenticated attacker run code on the server over HTTP, with no login required. Google's Mandiant says the bug was exploited from May 27 to June 9, before any advisory existed, and notified more than 100 affected organizations, 68 percent of them universities. The exposed component is the Environment Management Hub. Affected versions are PeopleTools 8.61 and 8.62; a full patch is still pending.
The critical Ivanti Sentry flaw covered yesterday is now under active attack, with researchers reporting compromised gateways within about 24 hours of the patch and public patch analysis. CVE-2026-10520, rated a perfect 10, is an OS command injection in an internal configuration API that accepts commands from anyone who can reach it over the internet, granting remote code execution as root with no login. A second flaw, CVE-2026-10523, lets attackers create their own admin accounts. With exploitation confirmed and detection tooling public, the time to patch has effectively run out for internet-exposed appliances. Ivanti released fixes earlier this week.
Fortinet has patched a critical flaw in FortiSandbox, the appliance that detonates suspicious files and feeds malware verdicts to the rest of a Fortinet security deployment. The bug (CVE-2026-25089, rated 9.8) is an OS command injection in the web interface that lets a remote, unauthenticated attacker run arbitrary commands by sending crafted HTTP requests. Compromising a sandbox is especially dangerous because attackers can both pivot deeper into the network and blind the very system meant to catch malware. Fixed versions are FortiSandbox 5.0.6 and 4.4.9, with matching updates for the Cloud and PaaS editions.
VulnCheck reports that attackers are actively exploiting an unpatched flaw in Langflow, a popular open-source platform for building AI applications. The bug (CVE-2026-5027, rated 8.8) is a path-traversal weakness: the file-upload endpoint does not clean the supplied filename, so an attacker can use directory-climbing sequences to write files anywhere on the server, a foothold that leads to remote code execution. Tenable, which found it, says the maintainers did not respond after three contact attempts in early 2026, and there is still no official fix. Early exploitation appears to be probing, with attackers writing harmless test files, but that usually precedes heavier attacks.
Ivanti has patched two critical flaws in Sentry, its mobile gateway appliance (formerly MobileIron Sentry) that sits in line between mobile devices and back-end systems like Exchange. The worst, CVE-2026-10520, rated a perfect 10, is an OS command injection in an internal configuration API that mistakenly accepts commands from anyone who can reach it over the internet, with no login, granting remote code execution as root. The second, CVE-2026-10523 (9.9), is an authentication bypass that lets attackers create their own admin accounts. No exploitation has been seen yet, but watchTowr has already published a patch analysis and a detection script, so the window is closing fast.
Researchers at Cyera have disclosed six vulnerabilities, collectively named Proto6, in protobuf.js, a JavaScript and TypeScript library for Google's Protocol Buffers data format that sees more than 50 million downloads a week. The flaws stem from the library trusting schema and metadata by default, so a single malicious schema or crafted payload can crash a service, inject code, or lead to remote code execution. Cyera demonstrated real attacks including poisoning CI/CD pipelines to leak build secrets and crashing WhatsApp automation bots. Because protobuf.js is embedded across cloud services, AI platforms, and build systems, the reach is broad. Fixed versions are 7.5.6 and 8.0.2.
Attackers are actively exploiting a flaw in LiteLLM, a widely used open-source gateway that routes requests to AI models, and CISA has added it to its known-exploited-vulnerabilities list. The bug (CVE-2026-42271) lets any authenticated user run commands on the host through test endpoints that spawn whatever command is supplied in the request. Chained with a separate Host-header bypass in the Starlette web framework (CVE-2026-48710), it becomes unauthenticated remote code execution, giving full control of the server, credential theft, and a foothold in connected AI infrastructure. Horizon3.ai has published a proof-of-concept. It follows a LiteLLM SQL injection flaw exploited within 36 hours last month.
Veeam has patched a critical flaw in Backup and Replication, one of the most widely deployed enterprise backup tools, that lets any authenticated low-privilege domain user run code remotely on the backup server. The bug (CVE-2026-44963, rated 9.4) only affects version 12 installations joined to an Active Directory domain; version 13, which uses a different architecture, is not affected, and workgroup setups are safe. No exploitation has been seen yet, but Veeam warns attackers often move quickly once patches reveal the flaw, and backup servers are a prime ransomware target because compromising them cripples recovery. The fix is build 12.3.2.4854.
Researchers at Bishop Fox have shown that three maximum-severity flaws Ubiquiti patched in May can be chained into a single attack that hands an unauthenticated attacker root access to UniFi OS Server with one crafted web request. Two flaws (CVE-2026-34908 and CVE-2026-34909) bypass the login gateway by abusing how the server reads encoded web addresses; the third (CVE-2026-34910) injects commands into the package-update feature, which runs with passwordless sudo, making escalation to root trivial. The flaws hit version 5.0.6 and earlier across widely used gear like UDM, UCG, and UNVR appliances. Bishop Fox released a free script to check for exposure.
TrustStrike Labs