How to identify AI created content and its cybersecurity implications

AI has completely changed how content is created, even this blog has been supported by AI. You’ll see that I use the word “supported”, this is an important distinction, we utilise AI to help us create content, but there is always a “human in the loop” (thank you Mr Nick Ellis for that phrase). 🤣

With the help of AI, organisations and individuals are able to create content at speed, whether its blogs, images, emails or social media posts. The positive is that it creates efficiency and greater productivity, but as the technology improves, a key challenge is introduced – how do you distinguish between AI content and authentic content. This is critical in an individuals ability to trust what they are viewing, especially in the growth of misinformation and cybersecurity threats.

Why does it matter?

AI content is easily and rapidly created, and for the “naughty” individuals of the world, it allows them to create malicious content that is harder to identify. For example:

• Phishing attacks – Once upon a time dodgy emails were littered with spelling errors, but AI now gives them content that is professionally toned and grammatically correct.
• Deepfakes – The cybercrims are now able to generate videos using an individuals face, whether its Donny Osmond (watch BBC’s Scam Interceptors) asking for money or its your own face for identify theft or fraud.
• Social media – AI is able to create huge swathes of content that is able to manipulate opinions or spread disinformation – we have seen this in political campaigns, recent asylum seeker cases, but also the benign – Lewis Hamilton and his big move back to Mercedes from Ferrari! 😕
• SEO blogs – This type of content is created to mislead readers, but also impact the results of search engines and the AI that is summarising content for you – AI overviews on Google search. Its like AI tricking AI, to trick you!!

Understanding how to identify AI created content is essential, so that you are able to identify what is real and what is not.

Signs of AI generated text content

• Generic content – “AI often produces text that lacks personal anecdotes or nuanced opinions.” – this is literally AI telling and showing you that it is lacking personality!
• Fabricated stats – AI has been known to fabricate its stats, it once gave me a percentage for a piece of content, I looked at its sources and they didn’t match!
• Em dash – AI love to use an em dash within sentences! ” — “

For different text content forms, there are additional things to look out for:

• Social media – Posts might have high engagement, but not a real conversation, and multiple posts from different sources will use the same phraseology, making them quite repetitive.
• Emails – Generic greetings and closing, emails can also seem too formal, and also emails are driving towards an action – always hover over links and question why am I getting this email.
• Blogs – Check the meta data for clues, this could be missing, the language may not reflect the content and the author may not be credible.

Detecting AI images and videos

• Hands – This is a classic, AI has had lots of trouble get hands right, but it is getting better in images. For videos, where deepfakes are being used, does the hand cut across the face, AI can struggle to fill these gaps in live videos.
• Shadows – Where is the light source? AI doesn’t always place the shadows in the right place and can be inconsistent.
• Meta data – AI generated images may not have any data, where as original images do.
• Reverse image search – stock or original images can sometimes be found in multiple place, whereas AI images may not exist in any database.
• Chaos – AI videos, can include surreal and explosive content, watch where people move or converge, AI can make people appear or disappear – just like magic! 🪄

Protect yourself and your organisation from AI

AI generated content is here to stay, and its benefits are undeniable. However, we need to stay vigilant and aware to protect ourselves and our organisations. Key ways to do this are:

• Utilise AI detection tools – GPTZero, Originality.ai, Grammarly are just a few of the tools available.
• Verify sources – Check the data sources.
• Training – Train your team to recognise phishing attempts and content that is suspicious.

Can AI be good for my organisation?

Yes, to put it simply! We utilise more than just AI, we couple it with automation, our aim is to take boring, repetitive tasks and utilise AI and automation tools to free a person of this burden. The key benefits for a person, are that they get to focus on high value work for their organisation, improving motivation, work satisfaction and increasing their time for growth. And a business gains, increased productivity, individuals that can focus on work that generates business growth and save money on how they grow their team.

Find out how Future Workplace could help your organisation save streamline operations, reduce manual effort and unlock smarter ways of working.

---