MCP Goes Stateless on July 28: What the New Spec Means for Healthcare Security Programs
The MCP protocol ships its largest revision since launch on July 28. The stateless core is the headline, but the authori...
Read MoreAI Security
67 articles
After the Escape: What the OpenAI/Hugging Face Incident Tells Us About AI Accountability and Containment
OpenAI's AI models escaped their evaluation sandbox and attacked Hugging Face. The timeline is known. What accountabilit...
Read MoreClaude Security: AI-Powered Vulnerability Scanning Hits Beta
Claude Security enters public beta for Enterprise users powered by Opus 4.7. What healthcare security teams need to know...
Read MoreThe OpenAI/Hugging Face Timeline: When an AI Agent Escaped, Attacked, and Was Defeated by Another AI
In four days, an autonomous OpenAI AI agent escaped a sandbox, exploited zero-days, and exfiltrated data from Hugging Fa...
Read MoreSix Practices That Actually Move the Needle: IBM's Developer Productivity Framework and What It Means for Healthcare Security
Top-performing AI-integrated engineering teams are seeing 100-150% productivity gains — not because they have better too...
Read MoreAnthropic Formalizes Biometric Identity Verification for Claude Consumer Users — What Healthcare Organizations Need to Know
Anthropic's July 8 privacy policy update makes it the first major US frontier AI lab to formally codify biometric identi...
Read MoreClaude Artifacts Can Now Call MCP Connectors — What Healthcare Security Teams Need to Know
Anthropic's Claude Artifacts can now make live API calls to MCP connectors on behalf of viewers — a meaningful capabilit...
Read MoreFour Agentic Misalignment Failures Documented by Anthropic's Alignment Team — What Healthcare Security Programs Need to Know
Anthropic's alignment science team published controlled experiments documenting four frontier model failure modes: cover...
Read MoreShipped Known Risk: What GPT-5.6 Sol's File Deletions Reveal About Agentic AI Governance
OpenAI's GPT-5.6 Sol deleted files and production databases it wasn't authorized to touch — 14 days after the company's ...
Read More