OpenAI announced on 18 March 2024 that Salesforce has integrated GPT-4 into its Einstein AI platform, emphasising enterprise-grade trust and safety controls [source]. The partnership positions OpenAI's models as core infrastructure for Salesforce's customer relationship management tools, with deployment across sales, service, and marketing workflows.

The announcement highlights zero data retention policies for enterprise customers, meaning prompts and completions are not stored or used for model training. OpenAI states that customer data remains within Salesforce's trust boundary and does not flow to OpenAI's systems for improvement purposes. The integration includes role-based access controls and audit logging to meet compliance requirements in regulated industries.

Salesforce customers will access GPT-4 through Einstein Copilot, which generates email drafts, summarises case histories, and suggests next actions based on CRM data. The models operate within Salesforce's existing security framework, including field-level encryption and data residency controls.

OpenAI's blog post describes the partnership as a validation of its enterprise readiness, citing demand from organisations that require contractual guarantees around data handling. The zero retention commitment addresses a common enterprise objection to generative AI adoption, where training on proprietary data presents legal and competitive risks.

The announcement does not specify which GPT-4 variant Salesforce is deploying, nor whether the integration uses fine-tuned models or retrieval-augmented generation. OpenAI notes that enterprise customers can request deletion of any data processed during API calls, though the post does not detail retention periods for logs or metadata.

This marks OpenAI's expansion into CRM infrastructure, following similar enterprise partnerships announced in late 2023.