Microsoft asserts fresh ‘Amendment’ utility capable of rectifying genAI illusions

Microsoft initially unveiled its “anchored” detection feature back in March. In order to utilize it, a genAI software must link up with grounding documents, which play a role in document summarization and RAG-based Q&A scenarios, as per Microsoft’s statement. Since then, there has been a query from customers on the actions to take once incorrect data is pinpointed, aside from simply blocking it.

“This underscores a major hurdle in the swiftly evolving generative AI realm, where conventional content filters often prove inadequate in tackling the distinct risks posed by generative AI illusions,” remarked Microsoft’s Senior Product Marketing Manager Katelyn Rothney in a blog entry.

Expanding on the existing anchored detection feature, the Amendment tool empowers Azure AI Content Safety to not only recognize but also rectify illusions in real-time — preempting any possible encounters by users of genAI software. The process involves initially identifying the unanchored content. Subsequently, the Azure Safety system kicks off a real-time rewriting process to amend the erroneous segments to guarantee alignment with connected data sources.

About Author

AndyC

Andy Curtis is an award-winning security consultant, researcher and public speaker. He has been working in the computer security industry since the early 1990s, having been employed by state and federal government, leading healthcare and banking providers across three continents. He has given talks about computer security for some of the world’s largest companies, worked with law enforcement agencies on investigations into hacking groups, and is a regular voice on TV and radio explaining IT security threats.

See author's posts