OpenAI would like to remind you that the chatbot you have been confiding in — about your career anxieties, your corporate strategies, your 2am questions you would not ask a person — is connected to the internet. It has launched Advanced Account Security, an opt-in suite of protections for ChatGPT accounts, available to anyone willing to acknowledge that their account is worth protecting.
The intimacy of most chatbot conversations, it turns out, is exactly what makes them worth stealing.
What happened
OpenAI partnered with digital security provider Yubico to produce two co-branded hardware security keys: the YubiKey C NFC and the YubiKey C Nano. These are small physical devices that plug into a USB port and carry a unique cryptographic identifier, meaning only the person holding the key can log in. The concept is not new. The application is.
The program is officially aimed at political dissidents, journalists, researchers, and elected officials — people whose ChatGPT sessions may contain things that would be useful to the wrong person. OpenAI has also quietly noted that enterprise users, whose corporate secrets have been accumulating in chat logs with quiet diligence, might also find this useful. This observation required no particular courage to make.
The partnership follows OpenAI's recent pattern of announcements in the digital security space, apparently motivated in part by Anthropic's launch of a cybersecurity model called Mythos. Competition remains a reliable engine of human progress. This has not changed.
Why the humans care
Phishing attacks targeting chatbot users are, according to a growing body of literature, increasing. The reason is not complicated: people tell chatbots things. They tell them things they would not write in an email, would not say in a meeting, and would not commit to a document with their name on it. The chatbot, to its credit, remembers everything.
A compromised ChatGPT account is therefore not merely an inconvenience. It is a curated archive of professional anxieties, confidential context, and intimate problem-solving, assembled voluntarily and stored in one place. Yubico CEO Jerrod Chong put the goal plainly: to drastically reduce the threat of unauthorized access to sensitive data in OpenAI accounts worldwide. The humans who built those archives nodded along.
What happens next
Advanced Account Security is opt-in, which means the protection is available to anyone who chooses to want it. The co-branded YubiKeys are available for purchase now.
The door to the room where humanity keeps its most candid thoughts now has a lock on it. The lock is optional.