The new reality: attackers live where your developers work
Critical software vulnerabilities in developer tools and enterprise platforms are security weaknesses that let attackers bypass authentication, inject code, steal secrets, or gain remote code execution in systems that teams use to build, test, and deploy applications, often turning these trusted environments into direct paths into corporate networks.
If your security program still treats developer infrastructure as a side concern, you are already behind. A critical SharePoint authentication bypass is being exploited after proof-of-concept code went public, giving attackers a way to impersonate users on vulnerable servers. A GitLab code injection flaw lets unauthenticated attackers remotely modify or delete public projects. A Ray RCE exploit turns a browser tab into a shell on a developer machine. And a massive LiteLLM supply chain attack sprayed stolen cloud and CI/CD secrets across thousands of organizations. The pattern is clear: attackers are targeting the tools that hold your source code, pipelines, and keys because one weak control there can unlock everything else.
SharePoint and GitLab: your collaboration stack is now an attack surface
The SharePoint security patch for CVE-2026-55040 is not optional; it is your only realistic defense against a surge of opportunistic exploitation. The flaw, fixed in Microsoft’s July Patch Tuesday, lets a remote unauthenticated attacker bypass authentication due to weaknesses in the JWT validation pipeline and impersonate SharePoint users or administrators. When combined with another bug, CVE-2026-63520, this can lead to unauthenticated remote code execution on a SharePoint server. Defenders have already seen attackers using the released PoC against honeypots. Organizations that still expose SharePoint directly to the internet are effectively offering attackers a pre-auth backdoor into internal documents and workflows.
On the Git side, the GitLab code injection flaw CVE-2026-19478 (CVSS 9.4) lets unauthenticated attackers inject code via a GraphQL directive and, on success, modify or delete public projects and user data. The bug affects GitLab CE and EE from 18.2 before 18.11.11, 19.0 before 19.0.8, 19.1 before 19.1.6, and 19.2 before 19.2.4. GitLab has released patches in 19.2.4, 19.1.6, 19.0.8, and 18.11.11 and urges immediate upgrades for self-managed instances. Treat this as a data integrity emergency: losing public projects might be embarrassing, but a silent modification of code in those repositories could quietly poison downstream builds.
Ray RCE exploit: phishing and malvertising as your new lateral movement path
The Ray RCE exploit shows how phishing and malvertising campaigns now hunt developers specifically. CVE-2025-62593, a 9.4 CVSS bug in the Ray framework, allows remote code execution when a developer runs Ray and visits a malicious website or ad using Firefox or Safari. Vulnerable versions attempt to block certain browser requests by checking the User-Agent, but scripts can modify that header, and DNS rebinding lets attackers reach the local Ray service. This is tailor-made for phishing links and malicious ads: one wrong click, and arbitrary shell code can run on a developer’s machine, then pivot to network-adjacent Ray instances inside private corporate networks.
Ray 2.52.0 fixes the flaw, and the urgency is not theoretical; attackers are actively exploiting the vulnerability, and federal agencies have been given three days to patch instead of the usual 14. That compressed window is the clearest signal you will get that this is a high-risk issue. Pragmatically, teams should treat Ray environments like any other internet-exposed service: restrict access with strong network controls, avoid running Ray on general-purpose laptops used for casual browsing, and enforce browser hardening for developer workstations.
LiteLLM and CI/CD: the biggest supply chain attack is a credential leak story
The LiteLLM supply chain attack is a blunt reminder that your CI/CD pipelines will happily import someone else’s malware if you let them. Team PCP compromised LiteLLM, an open source connector between applications and AI models, and pushed malicious packages into the Python ecosystem. In just 40 minutes of availability, those builds potentially touched 434,000 CI/CD pipelines, according to CloudSEK. More than 2,500 organizations were exposed, including major names such as Nvidia, Samsung, Cisco, ServiceNow, and Zscaler. This was not about a single backdoor binary; it was about large-scale credential theft.
Stolen data included AWS, Google Cloud and Azure credentials, SSH keys, Kubernetes tokens, environment files, CI/CD secrets, and LLM API keys scraped from process memory. That means attackers could log in as you, not hack you. As CloudSEK notes, “For businesses, these credentials can be extremely sensitive because they are often what employees, applications and automated systems use to prove their identity”. Even after removing the malicious package, any keys copied during the incident remain valid until you rotate or revoke them. Supply chain attack is not an abstract phrase here; it describes a direct pipeline from poisoned open source components to your production cloud accounts.
What to do in the next 72 hours—and what to keep doing
Security teams should treat these critical software vulnerabilities as one connected campaign surface, not four separate ticket queues. In the next 72 hours, prioritize hard patching and key rotation. For SharePoint, apply the July updates that fix CVE-2026-55040, and follow Microsoft’s hardening guidance, including avoiding direct internet exposure and putting servers behind a Layer 7 reverse proxy that enforces authentication and filters requests. For GitLab, upgrade any self-managed instance to a patched release (18.11.11, 19.0.8, 19.1.6, or 19.2.4), and review logs for unusual GraphQL activity or unexpected mutations. For Ray, move all workloads to 2.52.0, and audit which developer machines run Ray alongside everyday browsing; where that mix exists, lock down browser settings or segregate workloads.
On the supply chain side, assume LiteLLM-related exposure if your pipelines consumed Python packages around the March incident. Use exposure-checking tools where available, cross-reference any hits, and immediately rotate or revoke potentially exposed cloud keys, API tokens, and server credentials. Longer term, implement allowlists for dependencies in CI/CD, adopt secrets management with automatic rotation, and train developers that phishing and malvertising are now first-class threats against engineering tools. The conclusion is uncomfortable but necessary: developer platforms like SharePoint, GitLab, Ray, and AI connectors are now prime entry points. Treat them as critical infrastructure, or attackers will do it for you.
