Is Copilot for Microsoft 365 a deceiving deceiver?

Copilot took a different approach by fabricating details as it progressed. It raised concerns about “incomplete data, inaccurate labeling, inconsistent structures, and duplicated entries” – issues that I had not mentioned (and did not exist). It pointed out non-existent issues such as “numerous entries with missing data for crucial variables like client identification, purchase date, and product type…, incorrect labels for certain variables including gender. Some values were labeled as M or F, while others were labeled as Male or Female.”

Every piece of information it provided was inaccurate.

It falsely claimed that data such as product costs were obsolete. It stated, “I have included a spreadsheet featuring instances of the data inaccuracies I discovered, along with the origins and dates of the data.” Such a spreadsheet was not real. I had no instances of data inaccuracies, and no origins and dates for the imaginary data.

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