ChatGPT used for Qantas CEO's email to millions affected by data breach

ChatGPT was used to create the Qantas CEO’s email sent to the millions of customers affected by the airline’s data breach.

Earlier this month, Qantas announced it had suffered a cyber attack that exposed details of nearly 6 million customers, including details like names, email addresses and contact information.

Soon after, an email was sent on behalf of Qantas CEO Vanessa Hudson to those affected customers: “I know this incident has been concerning, and I am deeply sorry for the uncertainty this has caused,” the email read.

Each email included details for each recipient, listing the types of data that were exposed using bullet points.

A bizarre decision behind these bullet points contained a tell-tale sign that OpenAI’s AI chatbot was used in Hudson’s letter.

Have a tip/leak/story idea? Get in touch using my secure email or messaging apps here.

Unusually, Qantas’ apology email’s bullet points were an embedded image file — specifically a .PNG file — and not the typical bullet point Unicode character. (This was first noticed by Mastodon user Adam Kent).

The choice to use an image file meant that these bullet points had metadata, which is information that comes with a file like when it was created, details about its author, et cetera.

This metadata not only contained a disclosure that the file was AI generated (by marking it with a “Created using Generative AI” code), it specifically signed it as being created and modified by GPT-4o, a model released by OpenAI in 2024.

This raised the question: if the bullet point was created using ChatGPT, what else in Hudson’s letter was also the product of AI?

A Qantas spokesperson confirmed by phone that ChatGPT had been used in the production of the letter, but claimed that it was only used for formatting and not for producing any of the text.

Generative AI tools have become increasingly popular in white collar jobs, with one survey suggesting that many workers are using it without the knowledge of their bosses.

This technology has also been adopted by cybercriminals for nefarious purposes, including speculation that AI voice deepfakes were used to hack Qantas in the first place.