MCP for agent-to-agent comms may be the riskiest protocol you've never heard of

Ars Technica security feed covering emerging AI-agent security risks, prompt injection and protocol trust gaps, exploitation of critical software vulnerabilities, ransomware and data breaches, social-engineering campaigns such as ClickFix and scareware, exploit kits, privileged AI assistants, cloud and account compromises, surveillance-device exposure, cryptographic threats, and defensive developments including quantum-safe TLS. The feed includes reports of active exploitation and large-scale compromises, but most entries do not provide specific CVE identifiers.