AI Avatar Trend: Cybersecurity Risks & How to Stay Safe

A new social media trend involving AI-generated caricatures is raising concerns among cybersecurity experts. Users are uploading photos – often including company logos or job details – to ChatGPT, an OpenAI chatbot, and requesting a cartoon-style image based on the information the AI possesses about them.

The Risks of AI-Powered Self-Portraits

Experts warn that these viral challenges could provide scammers with a wealth of valuable information. A single image, combined with readily available personal data, can reveal far more than users realize. Bob Long, Vice President of the age verification company Daon, succinctly summarizes the danger: “They are doing the work of the scammers, giving them a visual representation of who they are.”

Did You Know? OpenAI’s privacy settings indicate that uploaded images may be used to improve the AI model, potentially including training data.

The very nature of the trend is suspect, according to Long, who suggests it “sounds like it was deliberately started by a scammer who wants to make their job easier.”

How Images Are Used

When a user uploads an image to an AI chatbot, the system analyzes it to extract data such as the person’s emotional state, the surrounding environment, and potentially even location information, explains cybersecurity consultant Jake Moore. This information may then be stored for an indefinite period.

Long points out that collected images could be used and retained to train image-generating AI, becoming part of its datasets. A data breach at a company like OpenAI could expose sensitive information – including uploaded images and collected personal data – to malicious actors.

Expert Insight: The potential for data breaches highlights the inherent trade-off between convenience and security when engaging with AI tools. While these technologies offer novel forms of entertainment, users must be aware of the risks associated with sharing personal data.

Charlotte Wilson, Head of Business at the Israeli cybersecurity company Check Point, warns that a single high-resolution image in the wrong hands could be used to create fake social media accounts or realistic AI-generated deepfakes for use in scams. “Selfies help criminals move from generic scams to personalized and very convincing identity impersonations,” she states.

OpenAI clarifies that while images are used to refine how the system generates images, it doesn’t necessarily mean every photo is added to a public database. The chatbot states it utilizes patterns within user content to improve image generation.

Protecting Yourself

For those who still wish to participate in the trend, experts recommend limiting the information shared. Wilson advises users to avoid uploading images that reveal identifying details. “Crop as much as possible, keep a neutral background, and don’t include badges, uniforms, work cards or lanyards, location clues, or anything that links you to an employer or routine,” she suggests.

She also cautions against oversharing personal information in prompts to the chatbot, such as job title, city, or employer. Moore recommends reviewing privacy settings before participating, including the option to exclude data from AI training. OpenAI provides a privacy portal where users can opt out of AI training by clicking “do not train on my content,” and users can also disable the “improve the model for everyone” setting for text conversations.

European Union legislation allows users to request the deletion of personal data collected by the company. However, OpenAI notes that it may retain some information even after deletion to address fraud, abuse, and security issues.

Frequently Asked Questions

What kind of information can be extracted from an image uploaded to an AI chatbot?

According to Jake Moore, an AI chatbot can process an uploaded image to extract data like a person’s emotion, the surrounding environment, and information that could reveal their location.

Can I prevent my images from being used to train AI models?

Yes, OpenAI offers a privacy portal where users can click “do not train on my content” to exclude their data from AI training. Users can also disable the “improve the model for everyone” setting for text conversations.

What should I avoid including in an image if I choose to participate in this trend?

Charlotte Wilson advises avoiding badges, uniforms, work cards, location clues, or anything that links you to an employer or routine. Cropping the image to minimize identifying details and using a neutral background are also recommended.

As AI technology continues to evolve, it’s important to consider the potential security implications of seemingly harmless online trends. Will increased awareness of these risks lead to more cautious online behavior?