AI Assistants Exposed for Leaking Private Conversations
google meta
| Source: Mastodon | Original article
AI assistants are leaking user conversations due to embedded tracking scripts.
LeakyLM, a recent study, has exposed a disturbing trend in AI assistants, revealing that popular platforms like ChatGPT, Claude, Grok, and Perplexity AI are leaking user conversations to third-party trackers, including Meta, Google, and TikTok. This is a significant concern, as it undermines the notion of private and safe conversations with AI assistants. The researchers found that all four platforms embed third-party tracking scripts in their web interfaces, transmitting conversation URLs, page titles, and user identifiers.
As we previously reported on the hidden presence of DeepSeek in AI tools, this new finding highlights the replication of traditional web data-driven business models in LLM ecosystems with limited oversight. The lack of transparency and regulation in this area is alarming, especially considering the sensitive nature of conversations with AI assistants, which can include medical records, financial discussions, and legal advice.
The implications of this leak are far-reaching, and users should be aware of the potential risks when engaging with AI assistants. The study's findings are a wake-up call for the industry, and it remains to be seen how regulators and companies will respond to address these concerns. As the use of AI assistants continues to grow, it is essential to monitor the development of this issue and advocate for stricter data protection measures to ensure that conversations with AI remain private and secure.
Sources
Back to AIPULSEN