AI Child Safety: Protecting Children's Photos Online

In an increasingly digital world, safeguarding our children's privacy is paramount. The use of personal photos of children for AI training without consent raises serious concerns. This article explores the risks, ethical considerations, and practical steps parents can take to protect their children's digital footprint. We'll delve into data scraping, legal reforms, and expert insights to empower you to navigate the complexities of online child safety.

Unconsented Use of Children’s Images

A recent report by Human Rights Watch

has brought to light a concerning trend: Australian children's personal photos are being secretly used to train AI Tools. This means that images shared within family circles, intended for the eyes of loved ones, are being harvested and fed into artificial intelligence systems without the explicit consent of the children or their parents. This practice raises critical ethical questions about data privacy and the rights of children in the digital age.

The report highlights that a data set used for AI training included images from a Melbourne high school, featuring compromising photos of 50 girls. Such disclosures underscore the vulnerability of children’s personal information and the potential for misuse. SBS Hindi reported on this issue, emphasizing the need for greater awareness and protective measures.

The core issue Stems from data scraping, a process by which large quantities of data are collected from websites. Data scraping, also known as Web Scraping

, is an automated process used to extract data from websites. This data is often used for a variety of purposes, including training AI models. While data scraping itself isn't always malicious, it becomes problematic when personal and sensitive information is collected without consent. In the context of children's images, the implications are particularly alarming.

Many families post images of their children online without realizing the potential risks. These images, intended to share joyful moments with friends and family, can be scraped and used for purposes far removed from their original intent. The lack of consent transforms a personal act of sharing into a violation of privacy.

Understanding the scope of this issue requires acknowledging the volume of data that AI models need to learn. AI algorithms are trained on massive datasets to recognize Patterns, predict outcomes, and perform tasks. When these datasets include children's images obtained without permission, it can lead to several ethical and practical problems.

The situation has prompted calls for stronger regulatory frameworks. Mark Dreyfus, Attorney-General

, has introduced reforms in parliament aimed at curbing the creation of fake explicit images. These reforms reflect a growing recognition of the need for laws to protect individuals, especially children, from the misuse of their digital likeness. The core of these reforms is to criminalize the unauthorized creation and distribution of explicit material generated through digital alteration or AI.

Data Scraping: How It Works and Why It's a Threat

To understand how children’s photos end up in AI training sets, it’s essential to grasp the mechanics of data scraping. Data scraping, also known as web scraping, is an automated process used to extract large amounts of data from websites . It involves using bots or scripts to Gather information that is publicly available.

In the context of social media and other online platforms, data scraping can be used to Collect images, text, and other forms of content. This information is then compiled into datasets that can be used for various purposes, including training AI models. Here’s a breakdown of how data scraping works:

1. Identifying Target Websites: Data scrapers begin by identifying websites that contain the data they want to collect. These can include social media platforms, photo-sharing sites, and personal blogs.
2. Developing Scraping Scripts: Automated scripts or bots are created to navigate the target websites and extract specific types of data. These scripts are designed to mimic human browsing behavior to avoid detection.
3. Data Extraction: The scraping scripts extract the desired data from the website. This can include images, text, metadata, and other information.
4. Data Compilation: The extracted data is compiled into a structured format, such as a database or spreadsheet. This makes it easier to analyze and use.

Data scraping becomes a significant threat when it involves the unauthorized collection of personal and sensitive information. In the case of children’s photos, the risks include:

• Privacy Violations: Harvesting images without consent is a direct violation of privacy rights.
• Misuse of Images: Scraped images can be used for unethical or illegal purposes, such as creating deepfakes or training AI models for malicious activities.
• Data Security: The collection and storage of children’s data can lead to security breaches, exposing their personal information to unauthorized parties.

Understanding these risks is the first step toward implementing effective protective measures. Measures against this range from adding CAPTCHA checks to implementing Web Application Firewalls

.

It's important to state that not all scraping is bad. Market researchers may scrape data to better understand customer behavior .

The Legal Perspective: Attorney-General Mark Dreyfus

Attorney-General Mark Dreyfus has recently introduced critical legal reforms aimed at combating the misuse of digitally created explicit material. Speaking in parliament, Dreyfus emphasized the need to prohibit the creation and distribution of non-consensual explicit images. These reforms represent a significant step towards safeguarding individuals, especially children, from digital abuse.

These legal changes are designed to:

• Criminalize Non-Consensual Deepfakes: The reforms make it illegal to create and share sexually explicit material that has been digitally Altered or created without the consent of the individuals involved.
• Protect Vulnerable Groups: The legislation prioritizes the protection of children and other vulnerable groups who are often targeted by these types of abuse.
• Strengthen Privacy Rights: By criminalizing the misuse of digital likeness, the reforms aim to reinforce privacy rights and deter potential offenders.

Dreyfus’s initiative reflects a growing awareness of the need for laws to keep pace with technological advancements. As AI and digital tools become more sophisticated, it is crucial to have legal frameworks in place to prevent their misuse and protect the rights of individuals.

Technology and Child Rights: Insights from Hee Jung Han

Hee Jung Han, a Children's Rights in Technology Researcher from Human Rights Watch

, shared alarming findings with SBS regarding the type of personal images being collected and used to train AI.

Han emphasized that the photos were never intended for anyone other than family and friends. Even images from the first seconds of a child’s birth, the researcher noted, where the baby may be covered in amniotic fluid, can be used in AI training. A family would never expect such a personal photo to be scraped into AI used to make a deepfake .

Han further stated that children and families should have a right to post joyful moments of their lives, and they should be confident their photos will be protected under data privacy laws

.

Data Scraping Safeguards: Expert Tips from Harvinder Singh

Harvinder Singh, a professor at the Artificial Intelligence Research Center at Torrens University

, provides practical insights into how data scraping is carried out and what measures can be taken to prevent it. His expertise offers valuable guidance for both website administrators and individuals seeking to protect their online data.

Singh defines data scraping as an automated process that can extract large amounts of data from websites. The process is not inherently illegal, but Singh warns it can be unethical if misused. To combat data scraping, Singh recommends :

• CAPTCHA Implementation: A CAPTCHA implementation prevents bots from extracting data. By adding CAPTCHA, a distinction can be made between a human user and automated bot. A CAPTCHA will ask for the user to identify numbers and letters, a task an automated bot will fail.
• Rate Limiting: In this measure, web servers can be limited from a specific IP address for a specific time period. If requests are made in excess, they can be blocked.
• IP Blocking: Specific IP addresses with suspicious activity can be easily identified and blocked through this measure.

He recommends a third method, the use of the * .htaccess website to apply blocks to suspicious behaviour.

Step 1: Review Your Social Media Settings

Begin by reviewing the privacy settings on all social media accounts where you share photos of your children. Ensure that your accounts are set to private, meaning only approved friends and family can view your content. While making your account private only keeps non-friends from seeing your posts, friends can still share content, and accounts can still be hacked, or the data otherwise scraped without permission.

• Facebook: Navigate to Settings & Privacy > Privacy Shortcuts > See More Privacy Settings. Adjust who can see your posts, limit the audience for posts you’ve shared with friends of friends or public, and review all posts and things you’re tagged in.
• Instagram: Go to Settings > Privacy > Account Privacy and toggle to Private Account.
• X (formerly Twitter): Go to Settings > Privacy and safety > Audience and tagging and select “Protect your Tweets.”

It’s not enough to set your account to private. Remove or avoid location tracking, disable facial recognition, and avoid using full names and birthdates.

Step 2: Limit Photo Sharing on Public Platforms

Be cautious about sharing photos on public platforms where anyone can view and download them. Even with privacy settings in place, there is always a risk that images can be scraped or shared without your consent.

Consider using private photo-sharing services or messaging apps with end-to-end encryption for sharing images with family and friends. These services offer greater control over who can access your content and provide additional security measures to protect against unauthorized access. A messaging app such as Signal [not mentioned in the audio but useful to list here], for instance, is free and provides end-to-end encryption, as does WhatsApp [not Mentioned in the audio but useful to list here].

Step 3: Implement CAPTCHA on Your Website

If you manage a website or blog where you share content, implement CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) on your forms and comment sections . CAPTCHA helps prevent bots from scraping data by requiring users to prove they are human before accessing certain features.

Implementing CAPTCHA is an easy way to prevent a large variety of automated scraping.

Step 4: Use Rate Limiting and IP Blocking

Website administrators can also implement rate limiting and IP blocking to protect against data scraping. Rate limiting restricts the number of requests that can be made from a specific IP address within a given time frame

, while IP blocking prevents access from known malicious IP addresses .

It is essential to know a suspicious user’s IP address before this measure can be used. This will require actively monitoring user activity.

Step 5: Enable Firewalls in Server Settings

Firewalls block any suspicious connections or users from accessing your website . Firewalls prevent data scrapers from accessing website info as well as stopping hacks and malware from being uploaded.