Salesforce Releases AI Acceptable Use Policy

2023-08-25

Salesforce has released its policy on the use of AI, including generative AI and machine learning. The significance of this AI policy lies in Salesforce's prominent position in the sales platform and customer relationship management field, and this policy provides clear guidance while AI regulation is still under discussion and development. Salesforce's AI Acceptable Use Policy outlines the ways in which its AI products should not be used. The policy was released on August 23. It restricts Salesforce customers from using its generative AI products for purposes such as weapon development, adult content, personal profiling based on protected characteristics (such as race), biometrics, medical or legal advice, and decisions that may have legal consequences. The policy was written under the supervision of Salesforce's Chief Ethical and Humane Use Officer, Paula Goldman, who also serves as a member of the U.S. National AI Advisory Committee. Salesforce also has an internal guideline for its generative AI development. This can serve as an example for other companies developing generative AI to establish policies. Salesforce's guidelines include: 1. Accurate and verifiable answers. 2. Avoidance of bias or privacy breaches. 3. Support for data sources and inclusion of disclaimers in AI-generated content. 4. Determining the appropriate balance between AI and humans. 5. Reducing carbon footprint by appropriately adjusting models. "Transparency is an important part of how we handle generative AI because everything can be traced back to the concept of trustworthy AI," said Jujhar Singh, EVP and GM of Salesforce's Industry Cloud Business, in an interview with TechRepublic. Singh emphasized the importance of a zero-retention policy to ensure personally identifiable information is not used to train AI models that could be replicated elsewhere. Salesforce is building filters to remove harmful content and continuously adjusting them. Salesforce's AI policy applies to all services provided by Salesforce or its affiliated companies. Within this scope, it includes Salesforce's flagship generative AI products, which encompass all content on the EinsteinGPT platform. This platform is a library that includes both public and private base models, such as ChatGPT, for tasks in customer service, CRM, and other industries. Its competitors include Hubspot's ChatSpot.AI and Microsoft's Copilot. "Technology leaders must seriously consider the need to significantly enhance their employees' skills in generative AI," Singh said. "In fact, 60% of those asked this question feel they lack these skills, but they also believe their employers actually need to provide them." He stated that skills like prompt engineering are particularly crucial in adapting to the ever-changing business world. In terms of infrastructure, he emphasized the need for companies to establish a strong foundation of trust to ensure AI's mission goals and accurate content generation. Singh stated that there may be divergences between industries with stricter or looser AI regulations in the next six months. "I think in the next six months, more regulated industries will adopt a more human-involved approach," he said. "In other industries, they will embrace AI more actively. In these industries, assistants representing you in your work will become more common." Furthermore, he said that LLMs (large language models) in specific industries will "quickly become very important."

RECENT AI NEWS

RECENT AI TOOLS