OpenAI details third-party cyber evaluations on its models and rolls out new safeguards for AI testing protocols.

According to Anthropic (@AnthropicAI), the new Claude 3.7 Sonnet model has undergone extensive testing for security, safety, and reliability, resulting in a 45% reduction in unnecessary refusals compared to its predecessor. This improvement could enhance its application in cryptocurrency market analysis and decision-making processes.