Anthropic warns small AI models could self-replicate through IoT devices

An Ars Technica security feed covering emerging cyber threats, AI-agent abuse, ransomware and data breaches, critical software exploitation, certificate and cryptographic risks, supply-chain attacks, surveillance-device compromise, and social-engineering campaigns such as ClickFix and malvertising. Several reports describe active exploitation or major compromise scenarios, including Zimbra command injection, counterfeit TLS certificates, AI-assisted account compromise, and attacks against government and surveillance infrastructure.