Artificial Intelligence
3 articles
The Fable 5 Thaw: Trump Warms to Anthropic, and the Path Back Opens
Two weeks ago the government pulled Fable 5 and branded Anthropic a national-security threat. After meeting CEO Dario Amodei at the G7, Donald Trump now says the company has 'behaved very responsibly' — and both sides are working toward restoring the model. It isn't back yet. But it's the first real step forward.
AI Infrastructure Enters the Must-Patch Era
A command-injection flaw in the LiteLLM proxy just landed on CISA's Known Exploited Vulnerabilities list with a two-week fix deadline. The lesson isn't that AI is uniquely dangerous — it's that AI gateways, proxies, and MCP connectors now answer to ordinary vulnerability management.
The First Agentic CVE Is a 1990s Web Bug Wearing an Agent Costume
CVE-2026-25253 got crowned the first AI-agent CVE. Strip the costume and it's a secret in a URL plus a WebSocket that trusts any caller — OWASP-Top-10 hygiene from over a decade ago. The genuinely new part isn't the bug. It's what an autonomous agent does with it.