Financial Times tests an AI chatbot trained on decades of its own articles

The Financial Times has introduced a new generative AI chatbot called Ask FT, designed to provide subscribers with curated natural language answers to a wide range of questions using decades of published information. This tool, akin to other generalized AI bots like ChatGPT and Copilot, draws on the outlet’s extensive archives to offer reliable responses. While the chatbot is currently available exclusively to a limited number of paid subscribers in the FT Professional tier aimed at business professionals and institutions, the outlet plans to keep refining its functionality.

Ask FT is powered by Claude, a large language model developed by Anthropic, although the technology used may evolve over time based on performance. The chatbot covers a spectrum of inquiries, from current events like Intel’s government funding under the CHIPS Act to more comprehensive topics such as the impact of cryptocurrency on the environment. It delves into the FT’s archives for detailed responses and provides citations for the referenced information.

While the tool generally offers accurate responses, some inconsistencies have been observed during testing, such as outdated data on candidates in the 2024 US presidential election. The team behind the chatbot is continuously fine-tuning its algorithms based on internal testing and user feedback, aiming to enhance accuracy and relevance.

Unlike previous AI tools deployed by other digital outlets, Ask FT appears to be more reliable and informative, delivering accurate answers to questions ranging from historical facts to recent developments. As the Financial Times refines and evaluates Ask FT further, the chatbot will remain in beta, with plans for wider availability to subscribers in the future.