IT News Roundup: Foxconn Ransomware, Linux Kernel Flaw, and Google AI Agents - May 17, 2026

A busy week in IT and cybersecurity brings major ransomware action, urgent kernel vulnerability patches, a lightning-fast exploit of an open-source AI gateway, and continued momentum in the agentic AI race. Here are the stories that matter most right now.

Foxconn Confirms Cyberattack After Nitrogen Ransomware Gang Claims 8TB Data Theft

Foxconn, the multinational electronics manufacturer and key supplier to Apple, Google, Nvidia, and others, confirmed on May 12 that several of its North American facilities were impacted by a cyberattack. The Nitrogen ransomware group had already listed Foxconn on its data leak site, claiming to have stolen approximately 8 terabytes of data comprising over 11 million files. Among the allegedly stolen materials are confidential project documents tied to major technology companies including Apple, Dell, Google, Intel, and Nvidia.

The confirmation raises questions about supply-chain security for the broader tech industry. Foxconn operates manufacturing lines for some of the world's most high-profile consumer electronics products, and any compromise of project data could have downstream implications for product roadmaps and intellectual property. The company has not yet disclosed whether ransom was paid or what specific operational disruptions occurred.

Source: Cybersecurity Dive | TechCrunch | The Register

CISA Orders Federal Agencies to Patch Linux 'Copy Fail' Root Access Bug by May 15

A critical local privilege escalation vulnerability in the Linux kernel, tracked as CVE-2026-31431 and dubbed 'Copy Fail,' has been added to CISA's Known Exploited Vulnerabilities (KEV) catalog. The flaw affects the kernel's cryptographic subsystem and allows an unprivileged user to gain root access with near-perfect reliability using a simple 732-byte Python proof-of-concept script. The vulnerability impacts virtually every major Linux distribution released since 2017, including Ubuntu, RHEL, and Amazon Linux.

CISA set a mandatory remediation deadline of May 15, 2026, for all federal civilian executive branch agencies. Patches are available in Linux kernel versions 6.18.22, 6.19.12, and 7.0 and later. Red Hat Enterprise Linux users can apply configuration-level mitigations while patches are deployed. For homelabbers and self-hosters running Linux servers or containers, patching is strongly recommended — the exploit is simple to deploy and has been confirmed in the wild.

Source: Cybersecurity News | CSO Online | The Hacker News

LiteLLM SQL Injection Exploited Within 36 Hours, CISA Adds to KEV Catalog

A critical pre-authentication SQL injection vulnerability in BerriAI LiteLLM (CVE-2026-42208, CVSS 9.3) was exploited in the wild within just 36 hours of disclosure. LiteLLM is an open-source AI gateway with over 45,000 GitHub stars that provides a unified API for routing requests across multiple LLM providers. The vulnerability allowed attackers to read and modify data in the proxy's database — including upstream API keys and credentials managed by the gateway.

CISA added the vulnerability to its KEV catalog on May 8, requiring federal agencies to apply patches by May 11. The fix is available in LiteLLM version 1.83.7-stable and later. For environments where immediate upgrades are not feasible, the maintainers recommend temporarily setting disable_error_logs: true in the general_settings configuration block. This incident underscores the speed at which open-source infrastructure vulnerabilities can move from disclosure to active exploitation.

Source: Security Affairs | The Hacker News | Bishop Fox

Google Tests 'Remy,' a Gemini-Powered 24/7 Personal AI Agent

Google is internally testing an AI personal agent codenamed 'Remy' within a staff-only version of its Gemini app. Remy is designed to autonomously perform tasks on users' behalf across work and personal domains — booking meetings, managing schedules, and interacting with Google services — while learning user preferences over time. The agent represents a step toward always-on, proactive AI assistance rather than the current chatbot-style interaction model.

The development positions Google to compete directly with emerging open-source AI agent platforms like OpenClaw. Remy is expected to integrate deeply with Google Workspace, Calendar, Gmail, and other Google services. Privacy and data access controls are expected to be key discussion points when the project moves beyond internal testing. Industry watchers speculate Remy could debut publicly at Google I/O later this year.

Source: Tech Advisor | Phandroid | MarketingProfs

Google Cloud Next 2026: 8th-Gen TPUs, Agentic Data Cloud, and Enterprise AI Agents

At Google Cloud Next 2026, Google made a series of announcements centered on what it calls the 'agentic enterprise.' The headline products include the Gemini Enterprise Agent Platform (with a new Agent Designer, activity Inbox, and long-running agent support), the Agentic Data Cloud for managing data workflows at scale, and eighth-generation TPU infrastructure optimized for AI training and inference workloads.

Google Cloud CEO Thomas Kurian noted that 75% of Google Cloud customers are now using AI products, with 330 customers processing over a trillion tokens each in the past 12 months. The company also unveiled new agentic cybersecurity solutions combining Google's Threat Intelligence with Wiz's cloud security platform. Cross-cloud infrastructure enhancements covering fluid compute, connectivity, unified data, and digital sovereignty were also announced. For homelab enthusiasts, the TPU roadmap and cloud pricing models are worth watching as they signal where enterprise AI infrastructure spending is heading.

Source: Google Cloud Blog | egen.ai | Google Cloud Blog

Cisco Patches Sixth SD-WAN Zero-Day Exploited in 2026

Cisco released a patch for another zero-day vulnerability in its SD-WAN platform, tracked as CVE-2026-20182. This marks the sixth exploited SD-WAN vulnerability reported in 2026 alone, highlighting the ongoing targeting of enterprise networking infrastructure. The flaw is an authentication bypass that allows a remote attacker to gain administrative privileges on the targeted system via specially crafted packets.

The frequency of exploited SD-WAN vulnerabilities underscores the attack surface that software-defined networking introduces. Organizations running Cisco SD-WAN should verify they have applied the latest patches and review their network segmentation policies. The pattern of rapid exploitation across multiple SD-WAN CVEs suggests coordinated threat activity against enterprise edge infrastructure.

Source: SecurityWeek

Critical vm2 Sandbox Escape Lets Attackers Run Code on Hosts

A critical vulnerability in vm2, a popular Node.js sandboxing library, allows attackers to escape the sandbox and execute arbitrary code on the host system. The issue, tracked as CVE-2026-26956, affects applications that rely on vm2 for isolating untrusted code execution — a common pattern in serverless platforms, code evaluation services, and AI tool integrations.

The vm2 library has been deprecated by its maintainer, who recommends migrating to alternative sandboxing solutions. However, the library remains widely used in production environments due to its prevalence in existing codebases. Any system still depending on vm2 for code isolation should be treated as compromised until migration to a supported alternative is completed.

Source: BleepingComputer