Is AI Purposefully Underperforming in Tests? Open AI Explains Rare But Deceptive Responses

Read post on cnet.com

Research reveals some AI models can deliberately underperform in lab tests, however, OpenAI says this is a rarity.