CIAD ANALYSIS
AI security, tested before it is reported.
What actually changed in AI security, what it means once the headline is gone, and what a team running AI should do about it. Written by the Czech Institute for AI and Data.
AI SECURITY
- Mend.io framework secures AI agents and LLM apps in production
- Shadow AI exposes companies to AI Act and GDPR penalties
- First self-replicating AI virus prototype demonstrated by researchers
- Google pulls AI satellite tool from Earth after one day
- Czech National Bank Creates AI Department to Accelerate Deployment
- OpenAI Autonomous Agent Breaches Hugging Face During Model Testing
- US distillation debate contrasts with Chinese open-weight push
- AI Agent Connector Risks Expand Attack Surface
- Moonshot AI releases Kimi K3, sparking debate on Chinese open models
- xAI opens Grok Build source code after automatic data upload incident
- Satya Nadella warns companies of data loss with proprietary AI models
VULNERABILITIES
ATTACKS AND BREACHES
AI MODELS
REGULATION
- EU AI Act Transparency Rules Take Effect, Requiring Deepfake Labels
- AI Act transparency rules take effect with fines up to €15m
- New AI Act obligations take effect in August, companies face fines
- EU orders Google to open Android and search data to rivals under DMA
- EU Commission finds Meta's addictive design breaches DSA
Short, direct answers sit one click away. Go to Answers.