EVA Tech BriefIT/DevOps

2026-05-26 - 7:30 PM PKT

Frontier LLMs

Claude Mythos - Anthropic's Restricted Security Model - May Be Coming to Claude Code

Anthropic appears to be preparing a public rollout of Claude Mythos, the restricted model that found 10,000+ critical software vulnerabilities in its first month. Previously limited to vetted security researchers, signs point to integration with Claude Code for automated vulnerability discovery during development.

ClickUp Lays Off Hundreds, Replacing Them With Thousands of AI Agents

Project management startup ClickUp has cut hundreds of employees and plans to replace them with thousands of AI agents. The nine-year-old startup says AI can now handle tasks previously requiring human workers, marking one of the most aggressive AI-for-humans swaps yet.

GPUs and Silicon

Cloud and Infra

Pakistan Tech

Pakistan Completes First Indigenous Earth Observation Satellite Constellation with EO-3

SUPARCO launched EO-3 aboard China's Long March-6 rocket, completing Pakistan's PRSC-EOS constellation (EO-1, EO-2, EO-3). The system enables indigenous climate monitoring, agricultural planning, and urban development analysis - Pakistan's first complete satellite constellation.

Security

KnowledgeDeliver LMS Zero-Day Exploited for Godzilla Web Shell and Cobalt Strike

CVE-2026-5426 (CVSS 7.5) in Digital Knowledge's KnowledgeDeliver LMS was exploited as a zero-day using hard-coded ASP.NET machine keys. Attackers leveraged ViewState deserialization to deploy Godzilla web shell and Cobalt Strike beacons. Popular in Japanese organizations.

7-Eleven Data Breach Exposes 185,000 People - ShinyHunters Gang Claims Responsibility

ShinyHunters extortion gang stole personal data including names, dates of birth, addresses, and Social Security numbers from 7-Eleven's systems in April. The breach is now confirmed via Have I Been Pwned and state government notifications.

Dev Tools

Novee Launches Agentic Fix - Pushes Pentest Findings Directly Into Claude, Copilot, and Cursor

AI pentest startup Novee ships Agentic Fix: validated exploit findings now push directly into developer AI coding agents (Claude Code, GitHub Copilot, Cursor) with remediation guidance generated from the same exploit context that found the vulnerability.

Uber President Says AI Spending Getting 'Harder to Justify' After Exhausting Annual Budget in 4 Months

After burning through its entire annual AI budget by April, Uber's COO says the company isn't seeing a clear connection between rising Claude Code token consumption and improved engineering output. Questions growing industry-wide about AI ROI.