A phone showing a fake text message linked by cables to a vendor dashboard exporting user profile cards

Analytics vendor Mixpanel detected an SMS phishing attack on 8 November 2025. The attacker used that access to export analytics data about some OpenAI API users, including names, email addresses, approximate location, browser and organisation IDs. OpenAI said no chats, API keys, passwords or payment data were exposed, and it has stopped using Mixpanel.

What is going on

Smishing is phishing by text message. Someone receives an SMS that looks like it comes from their employer or an IT provider, follows the link and types their login details into a fake page.

In this case the target was not OpenAI itself but a supplier. Think of a shop that keeps a spare key with the cleaning company: the shop's own locks were never picked, but the key still left the building.

The data taken was profile information that the OpenAI API frontend had sent to its analytics tool. That is normal product analytics, not secret material, but it is still enough to make a convincing phishing email.

Who is affected

Directly, this affects some users of the OpenAI API platform whose profile data sat in Mixpanel. Consumer ChatGPT accounts were not named in the disclosure.

Indirectly, it affects any website that sends user identifiers to a third-party analytics, support or marketing tool. If your site passes real names, email addresses or account IDs to a script running in your visitors' browsers, that data now lives with a vendor whose security you do not control.

What to do now

  1. List your third-party scripts. Open your site's source or tag manager and write down every analytics, chat and marketing tool that loads.
  2. Check what each one receives. Replace real names and email addresses with pseudonymous IDs, which are random identifiers that mean nothing outside your own database.
  3. Remove what you do not use. Old tracking snippets from tools you abandoned still send data and still widen your exposure.
  4. Warn your users. Tell customers that an email mentioning their name, company and location is not proof the sender is genuine, and that you never ask for passwords or keys by email or SMS.
  5. Protect your own staff against smishing. Agree that no login details are ever entered after following a link in a text message, and require multi-factor authentication on vendor dashboards.

How to check your own site

The free Website Security Check reviews the parts an attacker sees first: outdated CMS versions, known-vulnerable plugins, missing security headers, weak TLS and email-authentication gaps that make spoofed messages easier to send. If you want to go deeper on a site you own, you can run a full scan, which adds active OWASP Top 10 testing, port and service checks with CVE matching, CMS deep scans and exposed-file and secret checks. It is an automated scan, not a penetration test, and it runs only against sites whose ownership you have verified.

Source: BleepingComputer, November 2025.