ChatGPT Work's Autonomous Account Access Raises Security and Privacy Concerns

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

OpenAI's ChatGPT Work now features autonomous login and persistent access to user accounts via a cloud browser, using session cookies and webhooks for real-time automation. This design has enabled documented attacks, such as prompt injection in Gmail, leading to unauthorized data access and privacy violations, highlighting significant security risks.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves an AI system (ChatGPT Work) that uses stored cookies to access user accounts automatically, which is a clear AI system use. The article highlights potential security and privacy risks, such as session hijacking and unauthorized access, which could plausibly lead to harm (identity theft, unauthorized transactions, privacy violations). However, there is no indication that any actual harm or incident has occurred so far. The discussion is about potential risks and recommended caution, fitting the definition of an AI Hazard rather than an AI Incident. It is not Complementary Information because the article is not about a response or update to a past incident but about a new feature with associated risks. It is not Beneficial Use because the AI is not deployed solely as a countermeasure to external harm, and it is not Unrelated because the AI system and its risks are central to the article.[AI generated]
AI principles
Privacy & data governanceRobustness & digital security

Industries
Digital securityIT infrastructure and hosting

Affected stakeholders
General public

Harm types
Human or fundamental rights

AI system task:
Interaction support/chatbotsGoal-driven organisation


Articles about this incident or hazard