Artificial intelligencefromZDNET1 week agoAnthropic wants to stop AI models from turning evil - here's howNew research reveals persona vectors can help mitigate undesirable AI behavior like hallucinations or extreme agreeableness.