OpenAI Discloses Mixpanel Data Incident, Removes Vendor

OpenAI has disclosed that Mixpanel, a third-party analytics provider it used for web analytics on platform.openai.com, suffered an intrusion. Mixpanel discovered an attacker had exported a dataset from its systems on November 9, 2025, and shared the affected dataset with OpenAI on November 25, 2025. OpenAI stresses the incident did not involve unauthorised access to its own infrastructure.

Type of data involved

According to OpenAI, the exported Mixpanel dataset may have contained limited customer-identifiable and analytics information tied to users of the API platform. The company lists the categories that may have been affected:

• Name provided on the API account

• Email address associated with the API account

• Approximate coarse location (city/state/country) derived from browsers

• Operating system and browser details

• Referring websites

• Organization or user IDs associated with API accounts

OpenAI says that no chat content, API requests or responses, payment details, passwords, credentials, API keys, or government IDs were exposed.

OpenAI’s immediate response

OpenAI removed Mixpanel from production services, obtained the affected datasets for independent review, and began a security investigation alongside Mixpanel. The company says it is notifying impacted organisations, admins and users directly. OpenAI has also terminated its use of Mixpanel and is expanding security reviews and vendor requirements across its ecosystem.

Potential risks for impacted users

OpenAI warns the exposed fields could be weaponised in phishing or social-engineering campaigns. With names, email addresses and API metadata in play, attackers could craft credible messages that attempt to harvest credentials or trick recipients into revealing sensitive information. OpenAI recommends vigilance and highlights that users should treat unexpected messages with caution.

Practical steps OpenAI recommends

OpenAI repeats several standard security precautions for customers and developers:

• Treat unsolicited emails or messages cautiously, especially those with links or attachments.

• Verify messages claiming to be from OpenAI come from official OpenAI domains.

• Be aware that OpenAI will not request passwords, API keys, or verification codes via email, text, or chat.

• Enable multi-factor authentication (MFA) on accounts and enforce MFA at the single-sign-on layer for organisations.

OpenAI notes passwords, API keys and session tokens were not affected and therefore does not recommend blanket rotation as a direct response to this incident but stresses MFA and vigilance.

OpenAI says it is contacting affected users directly. Key follow-ups organisations and security teams will want resolved include the scope of affected accounts, whether any downstream logs or telemetry outside Mixpanel were exposed, what independent reviews found, and how OpenAI will harden vendor governance going forward.

OpenAI’s statement underscores a persistent risk in modern software stacks: third-party telemetry and analytics can create an unexpected attack surface. The company says it will elevate security requirements for partners and conduct broader vendor reviews, an explicit signal that enterprises and platform operators may increasingly demand stronger contractual and technical controls for analytics providers.

OpenAI characterises this as a Mixpanel incident that affected analytics data exported from Mixpanel’s environment; it says core product systems and user content were not compromised. Still, the event highlights the real-world consequences of vendor compromise: limited metadata can still enable sophisticated phishing, and organisations should treat partner telemetry as part of their threat model