Bridewell Study Finds LLM Privacy Policies Average 20-Minute Read

Bridewell assessed privacy policies from 20 LLMs. The Flesch Reading Ease Score, devised by Rudolf Flesch in 1948, measures readability; a score above 60 is considered good, while a score below that is not. The average score of 40.2 suggests the documents are dense with jargon that most people will not understand.

The reading burden is not uniform. The policy for Meta's Muse Spark is over 14,000 words long and takes almost an hour to read, TechRadar reported. As a rule of thumb, the more complex the LLM, such as Meta's Muse Spark or Moonshot AI's Kimi K, the longer the privacy policy takes to read.

The study also examined how LLMs use data. Bridewell found that 13 of the 20 models use inputs and outputs to train their models. Some LLMs offer users the option to opt out of training, while others do not, and even where an opt out is possible, it is not always clear how to action it. Around 75% of people admit to using an AI chatbot at least once a month, making it increasingly important to understand where questions and data queries go and how they are used.

Chris Linnell, Associate Director of Data Privacy at Bridewell, said: “It’s essential for users to fully understand how their data is being processed by LLMs, and businesses need clear internal guidance on what can and can't be shared, and ideally proper enterprise accounts with the right protections in place.” He added that employees may be at risk of sharing highly sensitive or confidential information that may end up being used to train LLMs.

The personal risk is not the only concern. Employees need to be aware of how they are using the technology for work, the study noted. Privacy policies are intended to provide that understanding, but their length and poor readability may leave many users agreeing to terms they have not reviewed.