AI summaries turn real news into nonsense, BBC finds

The BBC has conducted research into the accuracy of AI-generated news summaries following a significant error by Apple's AI service, which misrepresented a news story about a murder case. The study evaluated various AI assistants, including OpenAI's ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity, assessing their ability to provide accurate responses based on BBC news articles. The findings revealed that 51% of AI-generated answers contained significant issues, with 19% introducing factual errors and 13% misquoting or altering information from the original articles. Gemini performed the worst, with 34% of its responses judged problematic, while ChatGPT had the least at 15%. The BBC's Programme Director for Generative AI emphasized the potential value of AI when used responsibly but warned of the challenges it poses to the information ecosystem. The research highlighted the risks of misinformation and the potential erosion of public trust in factual reporting, raising concerns about the implications of AI in professional communication and critical thinking.

- The BBC's research found that over half of AI-generated news summaries contained significant inaccuracies.

- Apple's AI service was criticized for a major error in summarizing a news story, prompting the BBC's investigation.

- Gemini was identified as the least accurate AI assistant, with a high percentage of problematic responses.

- The study underscores the importance of responsible AI use to maintain public trust in news and information.

- Concerns were raised about the broader implications of AI on critical thinking and professional communication.