AS a senior figure in a tech company, Tom Coward is up-to-speed with digital innovations.
Yet, he found himself perplexed one day when a standard-looking email washed up in his inbox.
Taking no risks, Mr Coward – chief financial officer at risk digitisation and automation platform Cytora – went to Claude, Anthropic’s AI, for clarification, and the chatbot confirmed his vague suspicion that it was a fake.
This fairly innocuous-sounding experience, though, proved to be the spark for a new idea: why not streamline this makeshift vetting process for everyone?
After all, with AI continuing to ramp up in both scope and complexity, online scams have never been so dangerous.
“You’re not getting random emails from ‘Nigerian kings’ with loads of spelling mistakes anymore,” Mr Coward explained.
“The emails that are landing in my inbox, they look authentic – they’ve got AI that can do research about you, understand more about you, and target you.”
He decided to use Claude to actively counter this pernicious use of AI for scamming unsuspecting email users.
Incredibly, Mr Coward said it only took 45 minutes of conversation with the chatbot to move from “initial concept to working prototype”. He compared the “iterative process” of using Claude to a “choose your own destiny book”.
“I just have to say, ‘OK great, that’s the output’, and then try it, and if it works, great, feed that back to it, if it doesn’t, explain what’s gone wrong, then troubleshoot,” he said.
“I’m not an engineer or a developer, I can’t write code – but it was just incredible.”
His AI-generated prototype was quickly able to scan emails and write an “evaluation”, before grading the danger-of-scam on a scale from 1 to 10.
But Mr Coward soon found himself dreaming even bigger – it would take him a month to build an add-on allowing people to use the detector as a function within Gmail.
“That kind of thing just takes an additional bit of complexity, going from an initial working prototype to a fully-fledged self-sustained product”, he said.
In its current form, Mr Coward described the detector, guardmyemail.com, as more of a “training tool” than a security device.
He hopes his AI-invention will help the Island bolster its cyber-security.
“I want to work with the police to train it specifically to Jersey,” he said. “To train models to recognise Jersey specific threats, be aware when there are new threats, and feed that back to the police.”
He said his experience developing the product proves that AI is an accessible tool to be used and played around with by anyone.
“There’s no reason for anyone not to give it a go; it costs nothing, you can do it free.
“It’s not like the early days of Chat GPT – making it write a poem in a pirate voice, although that was pretty crazy – now it’s spitting this code that you can set with whatever weird or wild idea you’ve got, and try it out yourself.”
Despite his views on the benefits of AI, Mr Coward remains equally steadfast in his appreciation of the ‘human component’.
“Critical thinking, and being logical, and making business decisions – that has always been there, and that continues to be in there in this new world – but what’s different now is the actual execution of it.”
If you want a second opinion on an email’s authenticity, forward it to scan@guardmyemail.com
