ChatGPT Conversations Online: Data Erasure Efforts Fail – The Irish Times
Table of Contents
The ease with which sensitive facts shared with ChatGPT can become publicly accessible has been laid bare by a recent report, highlighting the risks for individuals and organisations alike. The Digital Digger inquiry revealed examples ranging from a multinational discussing building a hydroelectric facility, to an Egyptian using AI to critique their government, and a researcher documenting academic fraud – all potentially exposed through a flaw in ChatGPT’s settings.
ChatGPT’s Search Engine Visibility Issue
at the end of last month, it emerged that thousands of ChatGPT users inadvertently made their conversations visible to search engines like Google. This occurred because users were essentially ticking a box without fully understanding the implications.OpenAI quickly disabled the feature, described as a “short-lived experiment” by a senior executive, and “scrubbed” approximately 50,000 conversations.
Experts emphasize the difficulty of truly erasing information once it’s been made public online. The incident also underscores a broader issue: a lack of AI literacy among the growing number of people using these tools.
The Potential Consequences of Exposed AI Conversations
The consequences of this exposure extend far beyond mere embarrassment. Pradeep Sanyal, a San Francisco-based AI strategy advisor to company boards and CEOs, warns of important risks.
“Some posts reveal commercially sensitive information, legal strategies, political dissent in authoritarian contexts and confidential personal matters such as health conditions,” he said. “These could lead to reputational damage, competitive disadvantage, regulatory scrutiny or even personal safety risks depending on jurisdiction.”
Sanyal points to specific examples,such as a lawyer discussing strategies to displace indigenous communities at the lowest possible cost. “This is not merely embarrassing; it could have legal and ethical implications.”
AI and Data Privacy: Lessons from the Early Days of Email
Barry Scannell, an AI law and policy partner at William Fry solicitors and a member of the Government-appointed AI Advisory Council, expressed shock at the material made accessible through routine Google searches.
He noted the incident highlights the types of information people are sharing with AI tools – ranging from commercially sensitive business strategies to deeply personal disclosures, with some users treating AI as a therapist.
“Some of this reminds me of the early days of electronic interaction when some people were very flippant about what they put in emails and that was sometimes shown up in revelation processes,” Scannell said. “It is indeed certainly another reminder of the importance of companies having very clear processes and policies when it comes to their people using this technology.”
The incident serves as a crucial warning about the need for greater awareness and caution when interacting with AI platforms, and the importance of robust data privacy policies for organisations utilizing these technologies.