Open – anthropic cross -tests expose risks of jailbreak and abuse – which companies must add to GPT -5 assessments
Openai and Anthropic have tested each other and found that even if the reasoning models are aligning better on security, there are still risks.
[og_img]
About The Author
