Three critical threats target web apps, digital trust, and AI systems

Vulnerabilities in PHP web apps

Two critical vulnerabilities, identified as CVE-2026-109618 and CVE-2026-109619, have been discovered in the PHP programming language. PHP is a widely used tool for building and operating web applications across the internet. Because of its prevalence, these flaws create significant risks for many different types of websites and services.

The vulnerabilities carry a reported CVSS severity score of 9.2, which is a high rating for security risks. They could enable unauthorized users to run harmful scripts or peek into private application memory. This means an outsider could potentially take over a server or steal data without needing a login.

The advisory from Aquila I notes that versions older than 8.5.12, 8.4.27, 8.3.36, and 8.2.35 are affected. Organizations using these versions are urged to identify their internet-facing applications and apply security updates immediately. Security teams should prioritize these fixes based on how critical the affected systems are to the business.

The eMudhra certificate breach

A second major threat involves the reported compromise of eight intermediate certificate authority (CA) private keys from eMudhra. Certificate authorities are the entities that establish trust across digital communications by verifying website identities. If these keys are compromised, the entire chain of trust for the associated certificates is put at risk.

The breach was attributed to critically weak key generation during the creation of the keys. Some P-384 private keys had values below 256, which could make them recoverable almost instantly by an attacker. This weakness allows for the potential issuance of fraudulent certificates for websites.

As reported by Indian Startup Times, the consequences could include website impersonation and the interception of encrypted communications. Mozilla has already blocked the affected certificates in the Firefox browser. Organizations need to inspect their SSL/TLS certificates to see if they are linked to these compromised intermediate chains.

AI supply chain attack on LiteLLM

The third threat is a supply chain attack targeting LiteLLM, a tool used to manage interactions with large language models (LLMs). Attackers distributed malicious packages and Docker images designed to steal sensitive information. These included LLM API keys, cloud credentials, and various application secrets.

The attack specifically affected LiteLLM version 1.107.0 and certain Docker images released on October 21, 2026. Because LiteLLM is often used in AI infrastructure, a compromise here can grant attackers access to broader cloud resources. This demonstrates how AI systems can inherit security risks from the software components they rely on.

Security teams are advised to validate their deployed package versions against trusted sources and check for suspicious outbound network connections. They should also rotate any credentials that might have been exposed during the period of the attack. This incident highlights the growing need for constant monitoring of software dependencies in AI environments.

A unified approach to security

While these three issues affect different parts of the technology stack, they share a common theme of hidden infrastructure risk. Vulnerable web apps provide entry points, compromised certificates undermine trust, and malicious dependencies expose AI secrets. Addressing these requires more than checking a single piece of software at a time.

Organizations need a unified way to see their exposure across external assets, certificate trust, and software supply chains. Security teams must be able to translate these technical findings into clear actions quickly. This involves moving from isolated security checks to a coordinated response that covers the entire digital environment.

Aquila I suggests a multi-layered approach to handle these threats. This includes scanning for external vulnerabilities, performing certificate trust analysis, and conducting threat hunting within AI infrastructure. By connecting these different areas, companies can better prioritize which risks pose the most material danger to their operations.