30 Days Gen AI Risk Trial -Start Now
Skip to main content
Image Gen·ChatGPT Free (limited); Plus $20/mo; Pro $200/mo; API $0.04-$0.12 per image·openai.com

DALL·E 3

OpenAI's text-to-image model integrated into ChatGPT and available via API. Generates high-fidelity images from natural-language prompts with strong text rendering and prompt adherence.

Risk Score
Critical
8/10

Independent assessment across data handling, compliance, security and transparency.

Overview

DALL-E 3 is OpenAI's third-generation text-to-image model, available to consumers inside ChatGPT Free, Plus, and Pro, and to developers through the OpenAI Images API. It excels at rendering legible text within images, following long compositional prompts, and producing on-brand marketing and illustration assets at 1024x1024 up through 1792x1024 resolutions. For enterprise risk, the default consumer tier (ChatGPT Free/Plus) has model training enabled by default and must be toggled off under Data Controls. API and ChatGPT Enterprise/Team usage is excluded from training by contract. Generated images carry C2PA provenance metadata. Copyright indemnity is offered only on enterprise tiers, so shadow-AI use on personal accounts exposes employees to IP and confidentiality risks when they paste proprietary references.

Risk factors

2
  • Default model training enabled for consumer tiers.
  • Potential IP and confidentiality risks for enterprise users.

Recommendations

2
  • Advise enterprise users to use the API with contract exclusions.
  • Implement strict data governance policies for user-generated content.

Data handling

Storage
US-based OpenAI infrastructure on Azure; 30-day abuse-monitoring retention even when training is disabled.
Retention
Conversations retained until user deletion; deleted chats purged within 30 days; abuse logs kept 30 days minimum.
Training on inputs
Consumer ChatGPT Free/Plus: prompts and images used to train models by default; user must opt out. API, Team, Enterprise: excluded from training by default.