OpenAI says it disrupted a Cambodia-based criminal network that used ChatGPT to support investment, romance, gambling, and impersonation scams. The case highlights how general-purpose AI can be folded into coordinated fraud operations and the growing importance of platform-level abuse detection.
The Kubernetes release team has published an early look at features and maintenance changes planned for version 1.37. Operators should review the upcoming deprecations, removals, and replacements ahead of the release to reduce upgrade risk across production clusters.
Adobe has issued a security update for CVE-2026-48449, a maximum-severity authorization flaw in Campaign Classic. The vulnerability carries a CVSS score of 10.0 and could allow arbitrary code execution without user interaction, making prompt patching a priority for affected enterprises.
Arch Linux has temporarily disabled adoption of orphaned AUR packages following a wave of malicious package takeovers. The response is intended to limit supply-chain abuse while maintainers address weaknesses in the community repository workflow.
New research finds structured AI data-pipeline tasks score 10.9 points below comparable free-form coding work. DataFlow-Harness is designed to narrow that gap by evaluating and improving how coding agents handle ingestion, chunking, quality scoring, and filtering in enterprise RAG workflows.
Google removed a newly launched Earth AI feature one day after release following criticism that it could generate fake imagery over real maps. The rapid reversal underscores the challenge of shipping generative tools in products where users may mistake synthetic content for geographic evidence.