Beyond Borders: How VPNs are Shaping Global Internet Access
Examine the growing privacy crisis in the AI era and learn how to safeguard your personal data
May 2024
•
5 min read
Asad Ch.
The rapid advancement of AI, including the rise of large language models (LLMs) and chatbots, introduces significant privacy concerns. Questions arise about whether personal information is part of a model’s training data, if prompts are shared with law enforcement, and whether chatbots can link various online activities to produce detailed outputs for others. These issues highlight the pressing need to reassess privacy in the AI era.
Key Privacy Risks Posed by AI Systems
AI systems amplify many existing privacy risks from decades of internet commercialization and unchecked data collection. The primary difference now is the scale: AI systems are incredibly data-hungry and opaque, giving us even less control over what personal information is collected, how it is used, and how to correct or remove it. Today, avoiding digital surveillance while using online services is nearly impossible, and AI exacerbates this situation.
Additionally, there’s a growing risk of our data and AI tools being used for harmful purposes. Generative AI, trained on data scraped from the internet, may memorize personal information, including details about our families and friends. This data can enable targeted identity theft or fraud. Bad actors are already using AI for voice cloning, impersonating individuals to extort money over the phone.
Moreover, data we share or post for specific purposes, like resumes or photographs, can be repurposed without our consent for training AI systems. This repurposing can have serious civil rights implications. For example, AI tools used in hiring processes have shown bias, such as Amazon’s AI screening tool, which was found to discriminate against female candidates. Similarly, facial recognition technology, while intended to catch criminals, often misidentifies individuals, particularly black men, leading to false arrests due to biased training data.
Reassessing Data Privacy Approaches
Despite the extensive data collection we’ve grown accustomed to, it’s not too late to create stronger regulatory systems that require user consent for data collection and mandate the deletion of misused data. Currently, our online activities are tracked extensively, from web browsing to mobile app usage with GPS-enabled devices. This pervasive tracking stems from industry practices established decades ago when the internet was commercialized.
An ideal approach would be to ensure data collection only occurs with explicit user consent. When users browse online, their data should not be collected unless they actively choose to share it, such as by signing up for a service or creating an account. Data should not be considered public without clear user consent.
In the past, data privacy concerns were mostly related to online shopping, where users didn’t mind sharing their purchase habits. However, ubiquitous data collection now trains AI systems, impacting society and civil rights significantly. Default rules and practices around data collection can and should be revised.
Implementing Effective Data Privacy Solutions
Regulatory measures like data minimization and purpose limitation are crucial. These rules, which are integral to European privacy law (GDPR) and California’s privacy laws (CPPA), should also be a part of federal privacy legislation. However, operationalizing these rules effectively is challenging. Regulators must ensure that companies only collect the data necessary for their specific purposes, which can be difficult to assess, especially for companies that engage in diverse activities.
To better protect privacy, a shift from opt-out to opt-in data sharing is essential. Technologies like Apple’s App Tracking Transparency (ATT) and browser-based opt-out signals, such as Global Privacy Control, can help. These measures prevent third-party tracking and data sales, giving users more control over their personal information.
A supply chain approach to data privacy involves addressing both data input and output in AI systems. Ensuring that training data does not include personal information and setting guardrails to prevent the disclosure of personal data in outputs are critical steps. AI regulation must encompass the entire data ecosystem, including training data, to effectively protect privacy and reduce bias.
Finally, focusing solely on individual privacy rights is insufficient. Collective solutions, such as data intermediaries, offer a way to manage data rights at scale. These intermediaries can negotiate data rights on behalf of individuals, providing greater leverage and protection. While challenging to implement in the consumer space, data intermediaries are a promising approach to strengthening privacy in the AI era.
Behind every review is our team’s real-life, multi-device, cross-location VPN showdown. We’re all about the facts, not the hype. We’ve spent hours testing VPNs on every device we’ve got – to bring you advice that’s as genuine as it gets.
We are an alliance of passionate cybersecurity experts, united by a shared mission to disseminate our in-depth knowledge of digital security measures. Our commitment is to empower internet users by providing them with the essential tools and insights needed to navigate the online world safely. Through topVPNspot.com, we offer a platform where our collective experience in cybersecurity converges to guide users in protecting their digital presence effectively.