Over a quarter of companies have impletemented generative AI bans among their staff, according to a new survey by Cisco. The firm’s annual Data Privacy Benchmark Study of 2,600 privacy and security professionals in 12 countries also revealed that two-thirds of those polled have both imposed guardrails on what information can be entered into LLM-based systems or prohibited the use of specific applications.
“Over two-thirds of respondents indicated they were concerned about the risk of data being shared with competitors or with the public,” wrote Robert Waitman, a director at Cisco’s Privacy Center of Excellence, in a blog post about the survey. “Nonetheless, many of them have entered information that could be problematic, including non-public information about the company (48%).”
Respondents aware of data privacy concerns
Unsurprisingly, the survey revealed a strong and growing acquaintance with generative AI among respondents, with 55% saying that they were “very familiar” with the technology compared to 52% saying they were unfamiliar in the 2023 poll. 79% of those surveyed this year, meanwhile, said that they were extracting “significant or very significant value” from generative AI applications.
Even so, the study revealed persistent reservations among respondents about the ability to keep company data safe and secure when deploying such tools. These included worries that AI could endanger their organisation’s legal and intellectual property rights (69%), the leaking of proprietary data to competitors or the general public (68%) or that the outputs produced by the application could be inaccurate (68%). 92% of respondents, meanwhile, agreed that deploying generative AI required “new techniques to manage data and risks.”
Despite these concerns, many respondents freely admitted that they had inputted sensitive company data into generative AI applications. 62%, for example, had entered information about internal processes, while 48% had sought new insight from generative AI applications using non-public company data.
Generative AI bans becoming more common
It also appears that companies are beginning to adapt to these temptations by imposing guardrails on the use of generative AI. 27% of respondents said that their organisations had banned the technology altogether, though most companies seem happy to use ChatGPT-like systems provided they were deployed with certain restrictions. According to the survey, this can come in the form of data limitations (63%), tool restrictions (61%) or data verification requirements (36%).
These findings echo similar public announcements throughout 2023 by companies in retail, manufacturing and financial services that they were imposing guardrails on the internal use of generative AI. In May, Samsung banned its employees from using generative AI, while Amazon, Apple and Verizon imposed similar restrictions. These decisions were likely motivated by regulatory uncertainty as much as they were concerns about data privacy, generative AI expert Henry Ajder told Tech Monitor in October.
“They realise,” said Ajder, “that if they completely go all-in on generative AI in its current form, and in the current regulatory landscape, they could end up with them having to finish implementing something that is no longer compliant, or is going to be incredibly costly to maintain.”