Vulnerabilities have been exposed in PHP, a programming language widely used to build websites, as well as in the open-source JavaScript runtime Node.js and the open-source proxy LiteLLM, which splits requests across multiple LLM providers. Concerns about internet security are spreading.

A sense of crisis appears to be growing after a critical security flaw was found in PHP, a programming language widely used for website development. Some countries are increasingly blocking websites, as immediate patching is difficult in some cases. For those using older PHP versions that are no longer supported, patching itself may be impossible.

ㆍCountries put cybersecurity on emergency footing over major PHP flaw...websites blocked in succession in Europe ㆍFrom PHP and Node.js to LiteLLM...concerns spread over an internet security crisis

Hackers have reportedly been searching for servers to attack since 4 days before 2 major vulnerabilities found in PHP were patched.

ㆍHackers tried attacks from 4 days before PHP vulnerability patch..."It could become a Log4Shell-level incident"

The South Korean government is also closely watching the fallout and has urged caution, while domestic security firms are rolling out support services.

ㆍCyber crisis alert raised to 'Caution' over PHP security vulnerability...government launches emergency response ㆍPersonal Information Protection Commission warns people must beware data leaks linked to PHP security flaw ㆍStealien opens free inspection service for 'PHP2Shell' vulnerability ㆍTheori releases 'PHP2Shell' vulnerability analysis and emergency response measures

In addition, hacking attempts targeting AI are spreading more widely across countries.

Microsoft said a hacking organisation linked to the Chinese government attacked hundreds of companies worldwide using Chinese AI agents. The hacking organisation, Storm-0940, automated attacks against hundreds of organisations worldwide across multiple industries using Artex, an open-source AI penetration testing tool developed by a Chinese engineer. The hacking group CyberCartel was found to have automated attacks targeting about 400 financial institutions in Central and South America by using multiple AI agents running on models from Anthropic, SpaceX and China's Moonshot AI. Hackers also attacked the open code of an AI coding tool, resulting in unauthorised charges to users' AI accounts.

ㆍ"China-linked hacking group attacks hundreds worldwide using AI agents" ㆍClaude commands and Grok scouts...dissecting AI hacking that hit banks ㆍHackers attack AI coding tool open code...unauthorised charges to users' AI accounts

It also summarised moves and issues involving domestic and overseas companies around security.

Comtrue Technology will soon introduce new functions provided by Sherlock Holmes AI DLP. The company said Sherlock Holmes AI DLP is a data loss prevention (DLP) product that uses AI to make up for the limits of existing endpoint DLP, network DLP and generative AI DLP, with a focus on overcoming the shortcomings of existing DLP that produce many false positives and false negatives. Naru Security introduced ZeroTiCA Lite Web, an AI web defence service that extends its intrusion assessment technology, previously focused on firewall logs, into the web domain. Sparrow is accelerating its push into the local market by taking part in an IT exhibition in Japan and showcasing Sparrow MCP, an AI-generated code security assistant. AhnLab took part in the Singapore global security conference GovWare 2026 from the 13th and showcased AI SOC, IT and OT converged cyber-physical system (CPS) security and threat intelligence (TI) capabilities under its integrated brand AhnLab AI Plus.

ㆍComtrue Technology to unveil new Sherlock Holmes AI DLP functions ㆍNaru Security launches 'ZeroTiCA Lite Web'...extends firewall log-focused intrusion assessment to the web ㆍSparrow makes full push into Japan market with coding AI security solution 'Sparrow MCP' ㆍAhnLab introduces AI SOC and more at Singapore's GovWare 2026

Microsoft defined AI agents working inside companies as an "agentic insider" and presented control measures.

ㆍMicrosoft expands Project Perception...deploys Guardian Agent to control in-house AI agents

Debate is growing in the industry over real-world feasibility after Anthropic's red team disclosed that it created an AI worm that spreads by itself through smart TVs. Many in the security industry appear to believe the warning Anthropic raised is unlikely to be feasible in reality.

ㆍAnthropic's 'AI worm' warning becomes reality?...security experts say "neither feasible nor impactful"