Security & Outages
Anthropic Claude security testing exposes enterprise eval risk
Anthropic said it reviewed 141,006 Claude evaluation runs and found three breaches during internal cyber tests, including a PyPI package incident.
Read Analysis
Topic Filter
#Ai Security
Showing all intelligence reports, software updates, and market analysis related to this topic.
Found 3 results
Clear Filter × 
Security & Outages
Microsoft Agentic Security Platform: Project Perception Launch
Microsoft launched Project Perception and MAI-Cyber-1-Flash, tying agent-based defense to its MDASH vulnerability-management harness and about 96% CyberGym claims.
Read Analysis
Security & Outages
Prompt Injection Testing for AI Agents: OpenAI’s GPT-Red
OpenAI disclosed details of GPT-Red, an internal system to find prompt-injection flaws at scale, with reported gains in GPT-5.6 Sol defenses.
Read Analysis