Government workers warned against sharing classified information with ChatGPT, Daily Telegraph reports

• UK government workers were reportedly warned not to input classified information into AI chatbots.
• A document leaked to the Telegraph told them not to share anything "classified" or "sensitive."

UK government workers have been warned not to input any classified information into AI chatbots such as ChatGPT, The Telegraph has reported.

The paper was leaked a six-page document that contained a set of guidelines about how government officials should work with generative AI. It suggested that these programs could generate "biased" output, and misleading results to questions.

The Telegraph reported that the documents instructed workers to "never input information that is classified, sensitive or reveals the intent of the government (that may not be in the public domain) into any of these tools," and cautioned that "personal data" that could reveal identifying information about people should never be put into generative AI tool, the paper reported, per the Telegraph.

The guidelines flagged that a lot of the output created by chatbots could look "credible" — but is susceptible to bias and misinformation, as it could use sources that government officials wouldn't otherwise trust, the paper reported

The documents also encouraged workers to "explore" the emerging technology — which has exploded in popularity since 2022 — by asking themselves how the question will be used by the AI, how the tool can mislead and how it operates.

Insider had not received a response from the UK Civil Service as this article went live.

British Prime Minister Rishi Sunak has embraced the mission to make the UK a national AI powerhouse.

The Tokyo Metropolitan government announced last week that it would use ChatGPT to help streamline admin and clerical work, as well as assisting in tasks such as writing Q&As.

Romania has also pioneered the use of AI chatbots in government — having unveiled Ion, an AI advisor to its prime minister, in March. The bot's role is to analyze what Romanian citizens are saying on social media, and use the information to help government leaders shape policy decisions.