Breaking01 / AI
AI misinformation costs 3–169 times less to make than check
VEX-Bench assesses the burden of checking AI-generated misinformation across several dimensions.

Researchers reported on 10 Nov that AI-generated misinformation with a high verification burden cost 3–169 times less to produce than to check using AI agents. The comparison is a researcher-published figure.
Their paper introduces VEX-Bench as a benchmark for assessing the verification complexity of misinformation generated by large language models (LLMs). It measures the complexity perceived during screening, when content is selected for checking.
The benchmark covers 6 high-stakes domains and 60 real-world topics, according to the paper. These are researcher-published figures.
The researchers assessed checkability, potential harm and signals of source credibility, among other dimensions. Their VEX score combines success in eliciting misinformation with verification complexity.
The paper reported using AI as an evaluator and checking its assessments through agreement between evaluators and fact-checking agents. The researchers found that no single generation method led across all dimensions. They reported that content with a high verification burden often received priority during screening, consuming limited checking resources.
The researchers did not disclose a model-by-model breakdown of generation and verification costs in the abstract.