The paper Counter Turing Test CT^2: AI-Generated Text Detection is Not as Easy as You May Think – Introducing AI Detectability Index by Megha Chakraborty et al. discusses a new benchmark, Counter Turing Test (CT^2), which examines the effectiveness of AI-generated text detection (AGTD) methods. The study reveals the vulnerability of existing AGTD techniques and proposes the AI Detectability Index (ADI) to rank Large Language Models according to their ability to evade detection.

• CT^2 demonstrates the fragility of current AGTD methods.
• ADI measures the spectrum of detectability across various LLMs.
• Larger LLMs tend to be less detectable, posing greater challenges for text attribution.

The introduction of CT^2 and the ADI is a pivotal advancement in AGTD research, offering a framework for ongoing method evaluation as AI models grow more sophisticated. Such tools are essential for constructing fair and transparent AI governance and regulatory policies.