The ChatGPT desktop application for macOS has been found to forward SSH agent credentials to its Codex coding agent, potentially exposing user identity and keys when interacting with untrusted text. This occurs when a user's SSH configuration enables agent forwarding, allowing the AI agent to authenticate with any key loaded on the user's computer. Separately, Google's Bard blog post update indicates that data from connected apps may be used for training generative AI models and is subject to human review, raising similar concerns about data privacy and identity exposure. AI
IMPACT Raises concerns about the security and privacy implications of AI agents handling user credentials and data, potentially impacting user trust and adoption.
RANK_REASON The article discusses a security vulnerability and privacy concern related to the functionality of an existing AI product, rather than a new release or significant industry event.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →