Anthropic's updated Constitutional Classifiers reduce jailbreak success rates from 86% to 4.4%. This safety technology becomes critical as the EU AI Act's Article 50 transparency obligations remain set for August 2, 2026, despite extensions for high-risk systems.
AWS launched an agent marketplace with Anthropic featuring Claude 4 models like Claude Sonnet 4. These new commerce agents can increase purchase completion by 60% while providing developers with specialized tools for retail, travel, and telecom applications.
Businesses face compounding financial burdens due to a 50 percent price hike for Claude Sonnet 5. The rollout also faces challenges including coding performance degradation, agentic security threats like CVE-2025-49596, and complex regulatory compliance.
