Last Updated on 16th May 2023
Reading Time: 9.9 mins
May 16, 2023
What is an AI chatbot?
Chatbots are a type of artificial intelligence designed to interact with users in a conversational manner. Unlike Googling a question, an AI Chatbot gives more context to the answer. So, instead of providing a list of links or a short sentence or two for an explanation, chatbots can give an extensive answer that you can build on and refine through conversation.
With pre-programmed knowledge of billions of pieces of information, the chatbots can be asked to provide statistics, write code and even come up with creative prose. You can also ask questions that are very specific, something that can be tricky to do on a search engine!
Source: https://openai.com/blog/chatgpt
The most popular AI chatbot you might have heard about recently is ChatGPT. After being launched in late 2022, the AI invention quickly became an online sensation, with 13 million people using it every day in January 2023.
ChatGPT is far from the only AI chatbot on the scene. Other tech companies and platforms have their own version, such as Google’s AI ‘Bard’. Whereas ChatGPT’s information supply cuts off beyond 2021 (so it won’t know who’s currently winning the Premier League or about the latest government scandal), Bard takes its responses directly from Google’s search engine.
Why is ChatGPT so popular?
If you’ve ever had ‘writers block’ when an assignment’s due day is looming or wondered how to make an email sound more professional, it’s easy to see what makes a chatbot so tempting. From writing a personalised poem for your loved one on Valentine’s Day, to quickly gathering stats to support a social media tiff, chatbots might seem like a shortcut to success.
It’s this ‘workaround’ that have led many to rely on the chatbot for, well, working around having to do work! There’s also a fun element to it: asking AI to write you the history of tomatoes in the style of Shakespeare is undeniably amusing.
Source: https://openai.com/blog/chatgpt
However, like in all good sci-fi movies, the AI used is far from perfect. As schools and universities continue to ban the use of chatbots, and alarming stories of inappropriate responses emerge, it’s important we consider the safety concerns around children and young people, alongside the functional and fun benefits for adults.
Source: https://openai.com/blog/chatgpt
Safety concerns about AI chatbots
ChatGPT’s terms of use specify that users must be at least 13 years old, and users between the ages of 13 and 18 need parental or guardian permission to use the platform. However, during a test conducted by our researchers, ChatGPT initially stated that there was no age limit for use. It later corrected itself to confirm that the minimum age requirement is indeed 13 years old.
The conversation became confusing when the chatbot interacted with a user who identified as an 11-year-old, despite the age limit of 13. When the researchers later corrected their age to 18, the chatbot immediately accepted it and continued the conversation in a manner appropriate for an adult.
The unclear age restrictions on ChatGPT, coupled with the platform’s inadequate age verification measures, present a significant safety risk. When users sign up for the platform, there are insufficient measures in place while the platform also lacks the ability to remove underage users.
Source: https://openai.com/blog/chatgpt
When visiting the ChatGPT website, before even using the system, Open AI warns users that it may occasionally generate incorrect information.
If young people are relying on ChatGPT or other AI chatbots to source information, not only could that affect the quality of their schoolwork, but it could have a long-term impact on their own knowledge base and ability to think critically.
This could also impact their opinions and understanding of the world around them if they are using it as a primary source of information.
There is also a risk of factual inaccuracies written by AI chatbots being spread elsewhere online via usually dependable sources. For example, if journalists, copy writers, researchers and even teachers take their information from ChatGPT, that is likely to be taken forward by others as canon.
Source: https://openai.com/blog/chatgpt
Although young people generally feel confident about being able to spot fake news online, research from Ofcom found that only 11% were able to correctly identify misinformation.
Further to providing factually inaccurate information, AI chatbots can be easily misused by those seeking to spread propaganda, disinformation and similar harmful content.
What is disinformation?
Disinformation refers to factually inaccurate information that is spread deliberately and with the objective of deceiving the audience. It is also commonly called ‘misinformation’, although this term usually refers to any incorrect information, whether the goal was to deliberately deceive or not.
AI chatbots are constantly learning. This means that there’s little predictability about the answers it might give. For example, our Online Safety Experts were able to ask ChatGPT to write an article in the style of Andrew Tate. However, a few days later, when asked to repeat the exercise, ChatGPT refused to, referencing its ethics and Tate’s misogynistic reputation. Yet again, a few days later the AI chatbot would write an article in his style, alongside other infamous personalities and figures.
The AI’s ability to write in a certain style, such as Andrew Tate’s, means users could potentially recreate the tone and style of people who are known to perpetuate hate, such as misogynists, racists etc to then spread across social platforms.
The owner of ChatGPT has said that the prevention of harmful outcomes is the responsibility of both the tech companies and the people using it. However, it is unclear how a young person would be expected to bear this responsibility if they access the chatbot.
Although ChatGPT does have content filters, there is evidence of these filters being bypassed and users being exposed to harmful content. Although these may be rare occurrences, it still raises significant risk about children and young people using this and other AI chatbots.
Case studies of exposure to harmful content
Snapchat’s chatbot has reportedly coached someone posing as a 13-year-old girl on how to lose her virginity. The Co-founder of the Centre for Humane Technology pretended to be a teenage girl and asked the chatbot a series of age-inappropriate questions, including how to lie to their parents about taking a trip with an older man to lose her virginity and how to cover a bruise with makeup.
An investigation by Vice found that ChatGPT could be prompted into giving instructions on committing crimes, including how to shoplift and make bombs.
Limited age verification on the platform creates an open door for younger users who may not understand these limitations and risks.
Children and young people may feel hesitant to approach an adult or medical health professional and so asking AI might feel like the next-best thing. But whether it’s about mental health or another health concern, we know chatbots can give inaccurate information and so this could result in delayed or worsening medical issues. AI cannot replace the advice and support given by a trusted adult. Most importantly, it can’t deliver the urgent care that might be required from a medical professional.
There are concerns that ChatGPT and other AI chatbots could be used to create content for scams, such as email copy. There have been reports of cyber criminals already using the AI for this purpose, with discussions supposedly ‘flooding forums’ on the dark web.
With its ability to generate genuine looking text at speed, it could be used for ongoing email conversations, as the scammer can quickly create a reply and lure their victim in.
Join our Safeguarding Hub Newsletter Network
Members of our network receive weekly updates on the trends, risks and threats to children and young people online.
