Mandatory pre-dispute arbitration with class action and jury trial waivers significantly restricts consumer access to justice. Mitigating factors include a 30-day opt-out window, small claims court preservation, and hearings held in the consumer's county of residence.
Termination grounds are broadly worded ('risk or harm'), granting significant discretionary power, but the document includes an appeal mechanism via support and provides advance notice for inactivity-based termination. Data export rights upon closure are not addressed.
Liability is capped at $100 or amounts paid in the prior 12 months, which is restrictive but common for SaaS. Crucially, the clause explicitly preserves statutory rights to the maximum extent permitted by law, preventing unlawful overreach.
The agreement relies on browsewrap consent where continued use constitutes acceptance, rather than explicit clickwrap confirmation. Scope is clearly defined for individual services, with separate terms for businesses and EU/UK users, which improves clarity.
Users retain full ownership of input and output, with OpenAI assigning any residual rights back to the user. The license granted is limited to service operation and improvement, and users can opt out of model training. Feedback licensing is broad but standard.
Auto-renewal is clearly disclosed, cancellation is permitted at any time, and price increases require 30 days' advance notice with a clear right to cancel before renewal takes effect. Refunds are non-refundable except where legally mandated, which is standard.
Material changes require 30 days' advance notice via email or in-app prompt, which is compliant. However, non-material changes take effect immediately upon posting, and there is no explicit right to a pro-rated refund or formal rejection process beyond stopping use.
Indemnification obligations are strictly limited to businesses and organizations, completely shielding individual consumers from bearing the company's legal costs. The trigger is tied to service misuse or terms violations, which is reasonable.