98% of security leaders are concerned about the risks of giving third-party AI-based systems, including large language models, access to company data.