Get the latest tech news
Clio: A system for privacy-preserving insights into real-world AI use
A blog post describing Anthropic’s new system, Clio, for analyzing how people use AI while maintaining their privacy
In addition to training our language models to refuse harmful requests, we also use dedicated Trust and Safety enforcement systems to detect, block, and take action on activity that might violate our Usage Policy. For example, at this time we don't use Clio’s outputs for automated enforcement actions, and we extensively validate its performance across different data distributions—including testing across multiple languages, as we detail in our paper. And as we noted above, there are instances where Clio identified false positives (where it appeared there was activity violating our usage policy where there wasn’t) in our standard safety classifiers, potentially allowing us to interfere less in legitimate uses of the model.
Or read this on Hacker News
