Deepfake Awareness Training for Financial Organisations...
Deepfake fraud is becoming a practical risk for financial organisations, payment teams and SWIFT users. This article looks at how...
Read More
Artificial intelligence (AI) chatbots have become an increasingly popular source of information, with millions of people relying on them for everything from answering everyday questions to summarising news stories. However, a BBC investigation has highlighted an important concern: even without any deliberate manipulation, AI chatbots can distort and misrepresent the news they have been given.
The findings serve as a reminder that while AI is a powerful tool, it should not be treated as an infallible source of truth. As businesses and individuals continue to adopt AI into their daily workflows, understanding its limitations is becoming just as important as understanding its benefits.
The BBC tested four of the most widely used AI chatbots: OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini and Perplexity AI, by giving each of them access to BBC News content and then asking them a series of questions about current events. BBC journalists and subject-matter experts reviewed the resulting answers for accuracy.
The results were stark: more than half of the AI-generated answers contained “significant inaccuracies”, and clear factual errors appeared in roughly one in five responses. Microsoft’s Copilot and Google’s Gemini performed noticeably worse than ChatGPT and Perplexity.
Some of the specific errors uncovered included:
Notably, no one had to trick these systems with deceptive prompts. The chatbots were simply asked to summarise or answer questions about legitimate news content and still introduced distortions on their own.
Deborah Turness, CEO of BBC News and Current Affairs, did not mince words about the implications. She warned that AI companies are “playing with fire” and questioned how long it would be before an AI-distorted headline caused real-world harm. She called on tech firms to pull back their AI news summaries, pointing to Apple’s earlier decision to suspend its AI-generated news alerts after a series of high-profile errors, including one that wrongly claimed a murder suspect had shot himself and another that falsely reported that a tennis star had come out as gay.
The widespread adoption of generative AI means inaccurate information can spread faster and more convincingly than ever before.
Many users naturally assume that because an AI response is well written and confident, it must also be accurate. Unfortunately, confidence does not always equal correctness, and as the BBC’s findings show, this is not a rare edge case. It happened in over half of the tested responses.
For organisations, this creates several risks, including:
Businesses using AI to assist with content creation should therefore treat AI as a drafting tool rather than a final source of truth.
The errors uncovered by the BBC are a form of what is often called an AI “hallucination”, where a system produces information that sounds plausible but is actually false, even when working from accurate source material.
This can include:
Because these responses are delivered fluently and confidently, they can be difficult to catch without independent verification.
One of the clearest lessons from the BBC’s investigation is that fact-checking remains essential, even when AI is working from a trusted, verified source.
Good fact-checking habits include:
As AI tools become embedded in everyday business operations, organisations should establish clear guidelines for their use. Responsible AI use is not just about checking accuracy. It also means understanding the wider cyber security and privacy risks these tools can introduce.
For a closer look at how AI chatbots can create security risks for businesses and how to reduce them, read our article: AI Chatbots: When Help Becomes a Security Risk.
Best practices include:
AI can significantly improve productivity, but human judgement remains an essential safeguard.
AI chatbots are valuable tools that can assist with research, writing and problem-solving. However, the BBC’s investigation shows they are not immune to distorting even accurate, reputable source material, with no manipulation required.
The lesson for both individuals and organisations is straightforward: AI should complement critical thinking, not replace it. By verifying information, consulting reliable sources and maintaining healthy scepticism, users can reduce the risk of spreading misinformation while continuing to benefit from what AI has to offer.
At AJC, we help organisations understand and manage the risks created by emerging technologies, including AI tools used in everyday business activity.
Our team can support organisations with AI governance, cyber security awareness, risk management and practical guidance on using AI tools securely and responsibly.
As AI becomes more embedded in business workflows, organisations need clear policies, informed staff and appropriate oversight. AJC can help businesses strengthen their approach and reduce the risk of inaccurate, misleading or unsafe AI use.
Sources:
BBC News, https://www.bbc.co.uk/news/articles/c0m17d8827ko
https://deadline.com/2025/02/ai-chatbots-inaccurate-news-summaries-bbc-tech-1236285723/
AI Chatbots: When Help Becomes a Security Risk: https://www.ajollyconsulting.co.uk/ai-chatbots-when-help-becomes-a-security-risk/
Deepfake fraud is becoming a practical risk for financial organisations, payment teams and SWIFT users. This article looks at how...
Read More
Fraud controls should not only exist on paper. They need to be tested against realistic scenarios that reflect how criminals...
Read More
Compliance with recognised cyber security standards is an important part of managing risk, but it should not be mistaken for...
Read More