OpenAI and Anthropic cross-test models for hallucination and security issues

Jinse Finance reported that OpenAI and Anthropic have recently evaluated each other's models in order to identify issues that may have been missed during their own internal testing. Both companies stated on their respective blogs on Wednesday that this summer, they conducted safety tests on each other's publicly available AI models, examining whether the models exhibited hallucination tendencies and issues of so-called "misalignment," meaning the models did not operate as intended by their developers. These evaluations were completed before OpenAI launched GPT-5 and before Anthropic released Opus 4.1 in early August. Anthropic was founded by former OpenAI employees.