Online Safety Weekly Update - 17th September 2025

I hope you are keeping well and having a good week, albeit quite a change in the weather which is making my 6AM walk every morning less appealing!

It seems to be a week for chatbots this week as you’ll see by the updates below. This has been an increasing concern over the last couple of years and it’s continually increasing.

Updates

AI chatbots being used as therapists.
Fake celebrity chatbots – harmful content.
For parents – location sharing.
For parents – Roblox update

AI Chatbots Being Used as Therapists

At the beginning of September the NHS urged young people to stop using AI chatbots as a substitute for therapy (link HERE) due to the potential for harmful and dangerous advice. Urging people to stop using bots is one thing, and the NHS is absolutely right to warn people, but when it’s almost impossible to get the help and support you need people are going to turn elsewhere, and for many that elsewhere will be an AI chatbot.

The warning from the NHS comes after research found more than 17 million TikTok posts about using ChatGPT as a substitute and a YouGov poll which found around a third of 18-24 year olds said they would be comfortable discussing mental health issues with an AI chatbot instead of a real therapist. Personally I’ve found this with children Year 5 upwards, albeit at much lower numbers.

It’s vitally important we talk to children and young people about the incredible positives of AI and balance that with the shortcomings and risks. If you haven’t already seen it, Internet Matters released a fantastic report called ‘Me, Myself and AI’ which gives information on the ways in which children are using AI chatbots and will help with those important conversations. The link to the report is HERE.

Fake Celebrity Chatbots – Harmful Content

Following on from the update above, chatbots continue to hit the headlines with more and more research showing the very significant concerns. The Center for Countering Digital Hate (CCDH) released a big report in August (link HERE) where a large scale safety test on ChatGPT was carried out with some alarming findings related to mental health, eating disorders and substance abuse with over half of the 1200 responses from ChatGPT gave harmful advice.

Following on from that, in September Sky reported on some research carried out by ParentsTogether and Heat Initiative with the CharacterAI bot where, during 50 hours of testing accounts registered to children aged 13-17 they found almost 700 harmful/dangerous interactions. Some of the bots on CharacterAI replicate a real or imaginary real person (e.g. a movie star), therefore adding an extra layer of realism/complexity.

There is a significant parasocial relationship concern here: if children and young people form a relationship with the bot, and that bot is giving harmful/dangerous advice, the consequences don’t need explaining.

This is definitely one to be aware of and perhaps share with parents. The link to the Sky News article is HERE.

For Parents – Location Sharing

We all know that location sharing can be a really good thing, but only when that sharing is managed appropriately and only with very specific known persons, not to everyone. Thankfully, gone are the days when you uploaded a post onto social media your location was automatically included, but there are still location sharing features in some apps. As a timely reminder for parents to check their childs device the Canadian Cyber Tip Line has put together a very informative article which is useful for everyone, regardless of where you are in the world.

You can find the article HERE.

For Parents – Roblox Update

As the most popular game on the planet with children, Roblox gets a lot of coverage and understandably so. It’s an amazing platform to explore but it isn’t without its faults, some of which are very significant.

In April 2025 Revealing Reality carried out an investigation into Roblox due to serious concerns of what children were experiencing within the game (link HERE). In September they carried out a follow-up investigation to see what, if anything had changed.

In short, quite a lot has changed, such as:

Adjustments to content maturity ratings.
ID verification for certain games and experiences.
Stricter moderation
Enhanced parental controls.

These are good but, according to Revealing Reality, they don’t go far enough. There are still workarounds and very significant concerns. It’s worth having a read of the full report which you can find HERE.

Weekly Update to your Inbox

Online Safety Pro

Staff, parent and governor online safety training.

Annual staff, parent and governor training (pre-recorded videos).
Half-termly staff and parents updates (video for staff, PDF newsletter for parents.
12 month subscription – 1 single low cost per school (not per person).
All written and produced by Alan Mackenzie, an online safety specialist for over 17 years.
Discounts for multi-school or MAT purchases.