Indonesian Political, Business & Finance News

Beware of AI Chatbot Privacy: Your Conversations Are Not Entirely Secret

| | Source: Media Indonesia Translated from Indonesian | Technology
Beware of AI Chatbot Privacy: Your Conversations Are Not Entirely Secret
Image: Media Indonesia

Many users rely on AI chatbots for various purposes, ranging from health issues and romantic relationships to profound existential questions. However, behind this convenience lies a reality that is often ignored: these conversations are not entirely private.

A review by the Washington Post of public records found dozens of cases where chatbot transcripts were cited in court proceedings over the last two years. Often, these transcripts were found on users’ devices during searches by police or by opposing parties during the discovery phase of civil cases. In some instances, AI companies even reported suspicious content to law enforcement.

“Unless you are using a service that has a temporary chat feature or an incognito version, and you are using it in a browser that does not track you, the answer is no. You cannot be sure it is entirely private,” said Jen King, a privacy researcher at the Stanford Institute for Human-Centered Artificial Intelligence.

When you use chatbots such as Claude or ChatGPT, your messages are sent to a company’s data centre or servers for processing. If you log into a specific account, the chat history is saved and can be accessed again via any connected device.

In the legal cases reviewed, transcripts were often taken directly from the user’s physical device. In Missouri, a man allowed police to search his phone without a warrant. In Michigan, a lawyer voluntarily surrendered a client’s device as a form of transparency, which subsequently revealed AI conversation histories.

OpenAI and Anthropic state that if users delete conversations, the data is usually deleted from their internal servers within 30 days. However, there is an important exception:

Many companies use enterprise accounts that prevent AI from training its models using employee inputs. However, this does not mean employee activity is private from the company. Much like email or Slack, corporate accounts are managed by the employer. Administrators possess tools to view information, including chats and uploaded files.

OpenAI uses automated detection systems to identify harmful activities, such as violence or attempts at self-harm. The company will notify law enforcement if conversations indicate a real and imminent risk of harm to others. Anthropic also stated it would disclose limited information if it had a strong belief that it was necessary to prevent serious harm.

Both companies emphasise that they comply with legally valid subpoenas.

Ultimately, privacy in the digital world is very difficult to achieve. Although OpenAI and Anthropic state they do not sell data to third-party advertisers, full transparency remains a challenge because this technology requires access to your data to function.

More from 1 September 2026 →

Cite this article
Source: jawawa.id (https://jawawa.id/newsitem/beware-of-ai-chatbot-privacy-your-conversations-are-not-entirely-secret-1788249258)
Originally published by Media Indonesia (https://mediaindonesia.com/internasional/928298/waspada-privasi-chatbot-ai-percakapan-anda-tidak-sepenuhnya-rahasia), translated from Indonesian
Print