The risks that generative artificial intelligence models will leak private health information, enable cyberattacks, and produce incorrect answers should be addressed by the technology sector, according to guidance laid out Friday by US government scientists.
The National Institute of Standards and Technology, in a broad statement on AI risks, called on technology companies to carefully review the data they use to train their models to ensure it doesn’t include confidential information.
It also called on companies to:
- Make it harder for humans to direct models to create obscene images
- Allow people to remove their personal information from the data used ...