How to protect your privacy from generative AI

Generative AI has swept over the digital landscape with a tsunami of unprecedented innovation. Consumers across the globe are using applications like OpenAI’s ChatGPT, Bard, DALL-E, Midjourney, and DeepMind for content creation, ideation, and problem solving, or just simple fun. According to Nerdy Nav, the highest percentage of ChatGPT users are from the United States (15.22%), followed by India (6.32%).

Like any new technology though, generative AI too raises concerns about data privacy since it processes personal data and generates information that is potentially sensitive. The AI interactions can inadvertently be collecting personal data, such as names, addresses, and contact details of a user.

For instance, Google’s Bard has been facing flak for the possibility that it is trained on Gmail data of users. Also, according to Reuters, Google parent Alphabet has been cautioning its employees not to enter confidential information into chatbots, not excluding its own Bard.

The fact that OpenAI’s ChatGPT hasn’t been able to get too far in the European Union (EU), a regulation champion with strict data rules with only 3.98% user base of the total users globally, should alert us that generative AI should be handled with caution. In fact, the first known instance of a chatbot being blocked by a government order came in April when ChatGPT was banned in Italy over privacy concerns.

Why do these chatbots eye consumer data in the first place?

According to AI/ML developers, lack of data is the prime reason preventing the development of further AI models. Like the proverbial snake that swallows its own tail, generative AI is a great source of data for AI models, while data is the most significant component of making generative AI models.

In the end, AI is, after all, a technology, and as with any other technology, we can leverage it while keeping its disadvantages at bay with common sense.

As Mike Starr, CEO and Founder of trackd, a software company, says, “Despite the onslaught of breathless, and generally uninformed predictions about AI-driven impending doom, the best way to protect your privacy remains the same as always. Beware what you share on social media, protect your data with multi-factor authentication and don’t re-use passwords, and you’ll vastly reduce your chance of compromise, AI or no AI.”

Thankfully, there are ways to protect your privacy while using these useful generative AI tools.

1. Be Cautious with Personal Data

Avoid sharing sensitive personal information on platforms that use generative AI extensively. This includes details like your full name, address, phone number, or financial information.

“The power of generative AI algorithms means that we now need to think twice before sharing personal details, such as our full name, location, or personal photos online. We need to consider the potential implications of how this data can be used,” says Nate MacLeitch, Founder and CEO of QuickBlox, a cloud communication backend platform.

“Think about what you really want to share online and securely protect your online accounts with two-factor authentication and/or biometrics,” advises Jan Lunter, CEO and CTO of Innovatrics, a fingerprint recognition solution. “Generative AI doesn’t gather your identification data on purpose, this is more successfully done through phishing. Phishing can be aided by generative AI e.g. in generating images, deepfakes, professionally-sounding emails, or as a chatbot impersonating a real person. These attempts have skyrocketed in recent months.”

2. Practice Extra Caution with Office Work

If you are applying generative AI for office work or something confidential, be extra careful. Instances of ChatGPT leaking information are there.

In April, Samsung’s Semiconductor engineering team found this out the hard way. After developers put confidential code into ChatGPT twice, the chatbot absorbed it as training data to use for future responses to the public. As a solution, Samsung has been building its own AI system for its employees.

While we can’t all build ourselves a private AI system, it’s good practice to use generic identifiers or pseudonyms when interacting with generative AI models instead of sharing personal details. This helps maintain a level of anonymity and prevents the association of generated content with your real identity.

Reporting concerns or issues is important too. We can’t take these chatbots for granted, so if you happen to encounter any privacy concerns or suspect a breach of your private data while using a generative AI service, it’s imperative that you report the issue to the service provider and, if necessary, to relevant regulatory authorities.

3. Use a VPN

Using a Virtual Private Network (VPN) to anonymize user traffic can hide user location, preventing the AI from tracking the user across the web. A VPN can help by providing an encrypted connection as well as IP address anonymity.

When you connect to a VPN server, your real IP address is hidden, and you are assigned a temporary IP address from the VPN server’s location, which helps to obfuscate your true identity and location, making it harder for websites or services to track your online activities, including your usage of generative AI.

4. Review Privacy Policies

It’s a good idea to carefully read the privacy policies of the generative AI before using any platform. It’s important to understand how data will be collected, stored, and potentially used by the AI system. Look for transparency and clear information about data protection practices.

For example, in ChatGPT’s Terms of Use as of April 10, 2023, OpenAI says:

“When you use our non-API consumer services ChatGPT or DALL-E, we may use the data you provide us to improve our models.”

Though, OpenAI does offer an opt-out form so that ChatGPT or DALL-E doesn’t ingest the data.

It may sound tedious, but sometimes, looking for details on data collection, storage, retention, and sharing practices can surprise you. Choose services that prioritize user privacy, provide clear information about data handling, and have robust security measures in place.

5. Read Up on Data Storage & Retention

To understand data storage and retention of that data is another significant aspect. Find out about how long your data is stored and whether it is linked to your identity. Ideally, it’s better to opt for services that have a limited data retention policy and minimize the storage of user interactions.

Also, you can limit data retention by clearing your chat history. If the generative AI service allows it, it’s recommended to periodically clear your chat history or delete stored conversations. This will reduce the amount of data extracted from your interactions that’s likely to be stored or used for analysis.

These policies are also subject to updates and changes. So, stay updated and review notifications or announcements regarding data handling practices to ensure they align with your privacy preferences.

6. Check for Encryption & Secure Connections

Make sure that the generative AI service applies encryption to protect the data transmission between your device and the server. A good practice is to look for “https” in the URL and check if the service has security certificates.

For example, in the case of ChatGPT, OpenAI has not specified whether encryption is used for the transmission of data during the interaction with ChatGPT. However, ChatGPT primarily operates as a text-based AI model, and the conversation with ChatGPT occurs over an encrypted HTTPS connection between your device and OpenAI’s servers. This encryption helps protect the transmission of data.

It’s recommended to research the AI system’s track record in handling user data and prioritizing services with a strong commitment to privacy and data protection. Reading user reviews and experiences can also help. Any red flags or concerns raised by other users regarding data privacy can guide your choice.

7. Use Trusted & Reputable Services

Sticking to well-known and reputable generative AI platforms or providers is also a recommendation. Research the AI system’s track record in handling user data and prioritize services that have a strong commitment to privacy and data protection.

For people who are more technically sophisticated it is also worthwhile to explore privacy-focused tools and browser extensions that can help safeguard your online activities, says MacLeitch.

“These tools can block tracking scripts, prevent data collection, or enhance your privacy when browsing the internet.”

AI or No AI, Be Cautious

Generative AI is a new journey for many users in the exciting digital world where knowledge and content creation is at our fingertips. However, protecting privacy in this age is crucial. By understanding the technology, being cautious with personal data, reviewing privacy policies, and using trusted services, consumers can safeguard their privacy.

Further measures like strong passwords, regular updates, and VPN usage can further enhance privacy. Still, it’s necessary to keep yourself informed while adapting strategies as privacy risks evolve. To remain in control of your personal data in today’s digital landscape, it’s key to balance the benefits of generative AI with privacy protection.

By Navanwita Sachdev, The Tech Panda