FROM THE FRONTIER
ChatGPT can now monitor, flag, and report your conversations to law enforcement
Source: Made with Midjourney
OpenAI has unveiled a multi-layered monitoring system that’s actively scanning millions of ChatGPT conversations for harmful content — and in some cases reporting users to law enforcement. The disclosure comes in a recent blog post, where the company admitted its models have been implicated in multiple mental health episodes.
Here’s how the system works: The company says it uses automated filters to flag content that it deems to pose an “imminent threat of serious physical harm”, routing suspicious chats to what the company calls “specialized pipelines.” These chats land in front of a team of human reviewers, trained on OpenAI’s usage policies, who determine the best course of action, which could include escalating a chat to law enforcement.
The move raises serious data privacy concerns. The company is essentially building a content moderation system similar to social media platforms. But without transparency about false positive rates, appeal processes, or data retention policies, users may be left guessing whether their next prompt might end up in a police report.
via Superhuman