Many people and media outlets were tricked by a viral AI-generated image, but it highlights a very real danger

It is almost impossible to browse the Internet nowadays without stumbling into AI related topics or even an AI generated article, image or video. In the digital world, the artifices generated by Artificial Intelligence are more and more common becoming virtually indistinguishable for an untrained eye of an average user. The way this technology is utilized sparks numerous controversies and starts to break new boundaries. Not long ago, a new wave of concerns started regarding the use of AI in relation to learning. The counterfeit was so believable that multiple media outlets were caught on this deception. Are LLMs now capable of generating text using person’s own handwriting?

Is handwriting copying AI a reality or just another trick?

Handwriting is a personal attribute, a feature that is unique for every human being on earth. There can be similarities in two people’s way of encoding words onto a sheet of paper, yet there are always some minor differences that graphologists are trained to distinguish. However, the recent achievements of AI technology seem to completely nullify those variations and perfectly counterfeit a person’s handwriting style.

Source: X @immasiddx

Everything started with a post on X which showed a screenshot of a mathematical problem written in a notebook. The definite integral was solved by Gemini allegedly copying perfectly the handwriting of the person who prompted the LLM to solve it. As was later reported by the person who originally posted the image, multiple media outlets picked up the story reporting on concerns that such technology can bring.

In disbelief, another X user has asked to confirm whether this is indeed a real thing, but the author of the first post confirmed that it is a fake, and the whole image is AI generated, not only its solution. This was further explained in one of the threads on reddit, where it was pointed out that it is not a handwriting, but a font used by Nano Banana Pro, which is an image generation and editing model that’s developed by Google.

Browsing the Internet was never entirely safe, but in the current age of AI rise, being fooled is making this space even more treacherous. If, at any point, LLMs will become advanced enough to perfectly copy a person’s writing style, it can put into doubt many documents and open doors to easy forgeries that can be used in multiple methods of scam. Personal signature is still the most common way of confirming agreements, accepting regulations and so on. What will the future bring? Only the time will tell.

• tech
• AI / Artificial Intelligence