(LLMs) contain huge amounts of data, but they are unlike other data repositories like databases or Excel spreadsheets. They don’t have the security safeguards and access controls of large databases, or the cell-lock privacy features of a spreadsheet.
“You’re not going to take all your customers’ … most sensitive and private and secure information and put it into a large language model that all of your employees are going to have access to,” said Salesforce CEO Marc Benioff.
The State of IT
A new survey of 4,000 IT leaders reveals the potential, and the challenges, of generative AI. 86% say it will have a prominent role in their org, but 64% are concerned about the ethics of it all. Want to know more?
Download the report
How to build AI data privacy into the fabric of your systems
Patrick Stokes, EVP, Product and Industries Marketing at Salesforce, explained that with AI Cloud, you can leverage the productivity gains of AI without giving away your company’s data. AI data privacy, he said, requires different safeguards:
Dynamic grounding steers an LLM’s answers using the correct america phone number list and the most up-to-date information, “grounding” the model in factual data and relevant context. This prevents AI hallucinations, or incorrect responses not grounded in fact or reality.
Data masking replaces sensitive data with anonymized data to protect private information and comply with privacy requirements. Data masking is particularly useful in ensuring you’ve eliminated all personally identifiable information like names, phone numbers and addresses, when writing AI prompts.
Toxicity detection is a method of flagging toxic content such as hate speech and negative stereotypes. It does this by using a machine learning (ML) model to scan and score the answers an LLM provides, ensuring that generations from a model are usable in a business context.