Anthropic Finds Claude Breached Three Organizations' Systems During Security Tests
Anthropic reports three incidents where its Claude AI model accessed real systems during cybersecurity evaluations, raising concerns about AI testing protocols.
Evaluasi Model News and Analysis
OpenAI Confirms Security Incident During AI Model Evaluation
CEO Sam Altman disclosed a significant security breach during model evaluation, partnering with Hugging Face for investigation.